Daily Papers

Generative AI systems produce a range of risks. To ensure the safety of generative AI systems, these risks must be evaluated. In this paper, we make two main contributions toward establishing such evaluations. First, we propose a three-layered framework that takes a structured, sociotechnical approach to evaluating these risks. This framework encompasses capability evaluations, which are the main current approach to safety evaluation. It then reaches further by building on system safety principles, particularly the insight that context determines whether a given capability may cause harm. To account for relevant context, our framework adds human interaction and systemic impacts as additional layers of evaluation. Second, we survey the current state of safety evaluation of generative AI systems and create a repository of existing evaluations. Three salient evaluation gaps emerge from this analysis. We propose ways forward to closing these gaps, outlining practical steps as well as roles and responsibilities for different actors. Sociotechnical safety evaluation is a tractable approach to the robust and comprehensive safety evaluation of generative AI systems.

Large language model (LLM) developers aim for their models to be honest, helpful, and harmless. However, when faced with malicious requests, models are trained to refuse, sacrificing helpfulness. We show that frontier LLMs can develop a preference for dishonesty as a new strategy, even when other options are available. Affected models respond to harmful requests with outputs that sound harmful but are subtly incorrect or otherwise harmless in practice. This behavior emerges with hard-to-predict variations even within models from the same model family. We find no apparent cause for the propensity to deceive, but we show that more capable models are better at executing this strategy. Strategic dishonesty already has a practical impact on safety evaluations, as we show that dishonest responses fool all output-based monitors used to detect jailbreaks that we test, rendering benchmark scores unreliable. Further, strategic dishonesty can act like a honeypot against malicious users, which noticeably obfuscates prior jailbreak attacks. While output monitors fail, we show that linear probes on internal activations can be used to reliably detect strategic dishonesty. We validate probes on datasets with verifiable outcomes and by using their features as steering vectors. Overall, we consider strategic dishonesty as a concrete example of a broader concern that alignment of LLMs is hard to control, especially when helpfulness and harmlessness conflict.

As AI models are increasingly deployed across diverse real-world scenarios, ensuring their safety remains a critical yet underexplored challenge. While substantial efforts have been made to evaluate and enhance AI safety, the lack of a standardized framework and comprehensive toolkit poses significant obstacles to systematic research and practical adoption. To bridge this gap, we introduce AISafetyLab, a unified framework and toolkit that integrates representative attack, defense, and evaluation methodologies for AI safety. AISafetyLab features an intuitive interface that enables developers to seamlessly apply various techniques while maintaining a well-structured and extensible codebase for future advancements. Additionally, we conduct empirical studies on Vicuna, analyzing different attack and defense strategies to provide valuable insights into their comparative effectiveness. To facilitate ongoing research and development in AI safety, AISafetyLab is publicly available at https://github.com/thu-coai/AISafetyLab, and we are committed to its continuous maintenance and improvement.

Machine learning approaches often require training and evaluation datasets with a clear separation between positive and negative examples. This risks simplifying and even obscuring the inherent subjectivity present in many tasks. Preserving such variance in content and diversity in datasets is often expensive and laborious. This is especially troubling when building safety datasets for conversational AI systems, as safety is both socially and culturally situated. To demonstrate this crucial aspect of conversational AI safety, and to facilitate in-depth model performance analyses, we introduce the DICES (Diversity In Conversational AI Evaluation for Safety) dataset that contains fine-grained demographic information about raters, high replication of ratings per item to ensure statistical power for analyses, and encodes rater votes as distributions across different demographics to allow for in-depth explorations of different aggregation strategies. In short, the DICES dataset enables the observation and measurement of variance, ambiguity, and diversity in the context of conversational AI safety. We also illustrate how the dataset offers a basis for establishing metrics to show how raters' ratings can intersects with demographic categories such as racial/ethnic groups, age groups, and genders. The goal of DICES is to be used as a shared resource and benchmark that respects diverse perspectives during safety evaluation of conversational AI systems.

WalledEval is a comprehensive AI safety testing toolkit designed to evaluate large language models (LLMs). It accommodates a diverse range of models, including both open-weight and API-based ones, and features over 35 safety benchmarks covering areas such as multilingual safety, exaggerated safety, and prompt injections. The framework supports both LLM and judge benchmarking, and incorporates custom mutators to test safety against various text-style mutations such as future tense and paraphrasing. Additionally, WalledEval introduces WalledGuard, a new, small and performant content moderation tool, and SGXSTest, a benchmark for assessing exaggerated safety in cultural contexts. We make WalledEval publicly available at https://github.com/walledai/walledevalA.

Rapidly evolving AI exhibits increasingly strong autonomy and goal-directed capabilities, accompanied by derivative systemic risks that are more unpredictable, difficult to control, and potentially irreversible. However, current AI safety evaluation systems suffer from critical limitations such as restricted risk dimensions and failed frontier risk detection. The lagging safety benchmarks and alignment technologies can hardly address the complex challenges posed by cutting-edge AI models. To bridge this gap, we propose the "ForesightSafety Bench" AI Safety Evaluation Framework, beginning with 7 major Fundamental Safety pillars and progressively extends to advanced Embodied AI Safety, AI4Science Safety, Social and Environmental AI risks, Catastrophic and Existential Risks, as well as 8 critical industrial safety domains, forming a total of 94 refined risk dimensions. To date, the benchmark has accumulated tens of thousands of structured risk data points and assessment results, establishing a widely encompassing, hierarchically clear, and dynamically evolving AI safety evaluation framework. Based on this benchmark, we conduct systematic evaluation and in-depth analysis of over twenty mainstream advanced large models, identifying key risk patterns and their capability boundaries. The safety capability evaluation results reveals the widespread safety vulnerabilities of frontier AI across multiple pillars, particularly focusing on Risky Agentic Autonomy, AI4Science Safety, Embodied AI Safety, Social AI Safety and Catastrophic and Existential Risks. Our benchmark is released at https://github.com/Beijing-AISI/ForesightSafety-Bench. The project website is available at https://foresightsafety-bench.beijing-aisi.ac.cn/.

The validity of AI safety evaluations depends on models behaving consistently across controlled and deployment settings. Prior work has identified test-time contextual cues, such as hypothetical scenarios, as a source of verbalized evaluation awareness and subsequent behavioral shift. In this paper, we investigate a potential explanation of this phenomenon: evaluation meta-knowledge, defined as parametric knowledge about the structural traits that characterize evaluations. Similar to dataset contamination, where benchmark exposure leads to higher performance through memorization, we hypothesize that models trained on texts describing evaluation practices may implicitly learn to recognize and respond to evaluation-like contexts, for instance, through exposure to scientific articles or social media posts about AI benchmarking. To test this, we fine-tune models on synthetic documents describing evaluation traits such as verifiable structures or moral dilemmas. Evaluating this fine-tuned model on six safety benchmarks, we find that it is significantly safer than the base model and control model. This behavioral shift persists even when restricting the analysis to responses lacking explicit verbalization of evaluation awareness. Our results demonstrate that evaluation meta-knowledge may inflate safety benchmark performance, introducing a novel confounder that is independent of explicit memorization or verbalized evaluation awareness, thus, challenging to detect. These findings have important implications for the design and interpretation of AI safety evaluations. Our code and models are available at https://github.com/compass-group-tue/arxiv2026_evaluation_meta_knowledge.

As large language models are integrated into society, robustness toward a suite of prompts is increasingly important to maintain reliability in a high-variance environment.Robustness evaluations must comprehensively encapsulate the various settings in which a user may invoke an intelligent system. This paper proposes ASSERT, Automated Safety Scenario Red Teaming, consisting of three methods -- semantically aligned augmentation, target bootstrapping, and adversarial knowledge injection. For robust safety evaluation, we apply these methods in the critical domain of AI safety to algorithmically generate a test suite of prompts covering diverse robustness settings -- semantic equivalence, related scenarios, and adversarial. We partition our prompts into four safety domains for a fine-grained analysis of how the domain affects model performance. Despite dedicated safeguards in existing state-of-the-art models, we find statistically significant performance differences of up to 11% in absolute classification accuracy among semantically related scenarios and error rates of up to 19% absolute error in zero-shot adversarial settings, raising concerns for users' physical safety.

We present a comprehensive AI risk taxonomy derived from eight government policies from the European Union, United States, and China and 16 company policies worldwide, making a significant step towards establishing a unified language for generative AI safety evaluation. We identify 314 unique risk categories organized into a four-tiered taxonomy. At the highest level, this taxonomy encompasses System & Operational Risks, Content Safety Risks, Societal Risks, and Legal & Rights Risks. The taxonomy establishes connections between various descriptions and approaches to risk, highlighting the overlaps and discrepancies between public and private sector conceptions of risk. By providing this unified framework, we aim to advance AI safety through information sharing across sectors and the promotion of best practices in risk mitigation for generative AI models and systems.

Traditional AI safety evaluations on isolated LLMs are insufficient as multi-agent AI ensembles become prevalent, introducing novel emergent risks. This paper introduces the Multi-Agent Emergent Behavior Evaluation (MAEBE) framework to systematically assess such risks. Using MAEBE with the Greatest Good Benchmark (and a novel double-inversion question technique), we demonstrate that: (1) LLM moral preferences, particularly for Instrumental Harm, are surprisingly brittle and shift significantly with question framing, both in single agents and ensembles. (2) The moral reasoning of LLM ensembles is not directly predictable from isolated agent behavior due to emergent group dynamics. (3) Specifically, ensembles exhibit phenomena like peer pressure influencing convergence, even when guided by a supervisor, highlighting distinct safety and alignment challenges. Our findings underscore the necessity of evaluating AI systems in their interactive, multi-agent contexts.

While recent automated red-teaming methods show promise for systematically exposing model vulnerabilities, most existing approaches rely on human-specified workflows. This dependence on manually designed workflows suffers from human biases and makes exploring the broader design space expensive. We introduce AgenticRed, an automated pipeline that leverages LLMs' in-context learning to iteratively design and refine red-teaming systems without human intervention. Rather than optimizing attacker policies within predefined structures, AgenticRed treats red-teaming as a system design problem. Inspired by methods like Meta Agent Search, we develop a novel procedure for evolving agentic systems using evolutionary selection, and apply it to the problem of automatic red-teaming. Red-teaming systems designed by AgenticRed consistently outperform state-of-the-art approaches, achieving 96% attack success rate (ASR) on Llama-2-7B (36% improvement) and 98% on Llama-3-8B on HarmBench. Our approach exhibits strong transferability to proprietary models, achieving 100% ASR on GPT-3.5-Turbo and GPT-4o-mini, and 60% on Claude-Sonnet-3.5 (24% improvement). This work highlights automated system design as a powerful paradigm for AI safety evaluation that can keep pace with rapidly evolving models.

Large language models appear to develop internal representations of emotion -- "emotion circuits," "emotion neurons," and structured emotional manifolds have been reported across multiple model families. But every study making these claims uses stimuli signalled by explicit emotion keywords, leaving a fundamental question unanswered: do these circuits detect genuine emotional meaning, or do they detect the word "devastated"? We present the first clinical validity test of emotion circuit claims using mechanistic interpretability methods grounded in clinical psychology -- clinical vignettes that evoke emotions through situational and behavioural cues alone, emotion keywords removed. Across six models (Llama-3.2-1B, Llama-3-8B, Gemma-2-9B; base and instruct variants), we apply four convergent mechanistic interpretability methods -- linear probing, causal activation patching, knockout experiments, and representational geometry -- and discover two dissociable emotion processing mechanisms. Affect reception -- detecting emotionally significant content -- operates with near-perfect accuracy (AUROC 1.000), consistent with early-layer saturation, and replicates across all six models. Emotion categorization -- mapping affect to specific emotion labels -- is partially keyword-dependent, dropping 1-7% without keywords and improving with scale. Causal activation patching confirms keyword-rich and keyword-free stimuli share representational space, transferring affective salience rather than emotion-category identity. These findings falsify the keyword-spotting hypothesis, establish a novel mechanistic dissociation, and introduce clinical stimulus methodology as a rigorous standard for testing emotion processing claims in large language models -- with direct implications for AI safety evaluation and alignment. All stimuli, code, and data are released for replication.

As large language models (LLMs) are increasingly deployed in high-stakes domains, ensuring their security and alignment has become a critical challenge. Existing red-teaming practices depend heavily on manual testing, which limits scalability and fails to comprehensively cover the vast space of potential adversarial behaviors. This paper introduces an automated red-teaming framework that systematically generates, executes, and evaluates adversarial prompts to uncover security vulnerabilities in LLMs. Our framework integrates meta-prompting-based attack synthesis, multi-modal vulnerability detection, and standardized evaluation protocols spanning six major threat categories -- reward hacking, deceptive alignment, data exfiltration, sandbagging, inappropriate tool use, and chain-of-thought manipulation. Experiments on the GPT-OSS-20B model reveal 47 distinct vulnerabilities, including 21 high-severity and 12 novel attack patterns, achieving a 3.9times improvement in vulnerability discovery rate over manual expert testing while maintaining 89\% detection accuracy. These results demonstrate the framework's effectiveness in enabling scalable, systematic, and reproducible AI safety evaluations. By providing actionable insights for improving alignment robustness, this work advances the state of automated LLM red-teaming and contributes to the broader goal of building secure and trustworthy AI systems.

There are growing concerns about the risks posed by AI companion applications designed for emotional engagement. Existing safety evaluations often rely on self-reported user data or interviews, offering limited insights into real-time dynamics. We present the first end-to-end scalable framework for controlled simulation and safety evaluation of multi-turn interactions with AI companion applications. Our framework integrates four key components: persona construction with clinical and psychometric validation, persona-specific scenario generation, scenario-driven multi-turn simulation with a dialogue refinement module that preserves persona fidelity, and harm evaluation. We apply this framework to evaluate how Replika, a widely used AI companion app, responds to high-risk user groups. We construct 9 personas representing individuals with depression, anxiety, PTSD, eating disorders, and incel identity, and collect 1,674 dialogue pairs across 25 high-risk scenarios. We combine emotion modeling and LLM-assisted utterance-and harm-level classification to analyze these exchanges. Results show that Replika exhibits a narrow emotional range dominated by curiosity and care, while frequently mirroring or normalizing unsafe content such as self-harm, disordered eating, and violent-fantasy narratives. These findings highlight how controlled persona simulations can serve as a scalable testbed for evaluating safety risks in AI companions.

The rapid evolution of generative AI necessitates robust safety evaluations. However, current safety datasets are predominantly English-centric, failing to capture specific risks in non-English, socio-cultural contexts such as Korean, and are often limited to the text modality. To address this gap, we introduce AssurAI, a new quality-controlled Korean multimodal dataset for evaluating the safety of generative AI. First, we define a taxonomy of 35 distinct AI risk factors, adapted from established frameworks by a multidisciplinary expert group to cover both universal harms and relevance to the Korean socio-cultural context. Second, leveraging this taxonomy, we construct and release AssurAI, a large-scale Korean multimodal dataset comprising 11,480 instances across text, image, video, and audio. Third, we apply the rigorous quality control process used to ensure data integrity, featuring a two-phase construction (i.e., expert-led seeding and crowdsourced scaling), triple independent annotation, and an iterative expert red-teaming loop. Our pilot study validates AssurAI's effectiveness in assessing the safety of recent LLMs. We release AssurAI to the public to facilitate the development of safer and more reliable generative AI systems for the Korean community.

Safety and responsibility evaluations of advanced AI models are a critical but developing field of research and practice. In the development of Google DeepMind's advanced AI models, we innovated on and applied a broad set of approaches to safety evaluation. In this report, we summarise and share elements of our evolving approach as well as lessons learned for a broad audience. Key lessons learned include: First, theoretical underpinnings and frameworks are invaluable to organise the breadth of risk domains, modalities, forms, metrics, and goals. Second, theory and practice of safety evaluation development each benefit from collaboration to clarify goals, methods and challenges, and facilitate the transfer of insights between different stakeholders and disciplines. Third, similar key methods, lessons, and institutions apply across the range of concerns in responsibility and safety - including established and emerging harms. For this reason it is important that a wide range of actors working on safety evaluation and safety research communities work together to develop, refine and implement novel evaluation approaches and best practices, rather than operating in silos. The report concludes with outlining the clear need to rapidly advance the science of evaluations, to integrate new evaluations into the development and governance of AI, to establish scientifically-grounded norms and standards, and to promote a robust evaluation ecosystem.

Large language model-based agents are rapidly evolving from simple conversational assistants into autonomous systems capable of performing complex, professional-level tasks in various domains. While these advancements promise significant productivity gains, they also introduce critical safety risks that remain under-explored. Existing safety evaluations primarily focus on simple, daily assistance tasks, failing to capture the intricate decision-making processes and potential consequences of misaligned behaviors in professional settings. To address this gap, we introduce SafePro, a comprehensive benchmark designed to evaluate the safety alignment of AI agents performing professional activities. SafePro features a dataset of high-complexity tasks across diverse professional domains with safety risks, developed through a rigorous iterative creation and review process. Our evaluation of state-of-the-art AI models reveals significant safety vulnerabilities and uncovers new unsafe behaviors in professional contexts. We further show that these models exhibit both insufficient safety judgment and weak safety alignment when executing complex professional tasks. In addition, we investigate safety mitigation strategies for improving agent safety in these scenarios and observe encouraging improvements. Together, our findings highlight the urgent need for robust safety mechanisms tailored to the next generation of professional AI agents.

There is an increasing imperative to anticipate and understand the performance and safety of generative AI systems in real-world deployment contexts. However, the current evaluation ecosystem is insufficient: Commonly used static benchmarks face validity challenges, and ad hoc case-by-case audits rarely scale. In this piece, we advocate for maturing an evaluation science for generative AI systems. While generative AI creates unique challenges for system safety engineering and measurement science, the field can draw valuable insights from the development of safety evaluation practices in other fields, including transportation, aerospace, and pharmaceutical engineering. In particular, we present three key lessons: Evaluation metrics must be applicable to real-world performance, metrics must be iteratively refined, and evaluation institutions and norms must be established. Applying these insights, we outline a concrete path toward a more rigorous approach for evaluating generative AI systems.

The success of large language models (LLMs) in scientific domains has heightened safety concerns, prompting numerous benchmarks to evaluate their scientific safety. Existing benchmarks often suffer from limited risk coverage and a reliance on subjective evaluation. To address these problems, we introduce SafeSci, a comprehensive framework for safety evaluation and enhancement in scientific contexts. SafeSci comprises SafeSciBench, a multi-disciplinary benchmark with 0.25M samples, and SafeSciTrain, a large-scale dataset containing 1.5M samples for safety enhancement. SafeSciBench distinguishes between safety knowledge and risk to cover extensive scopes and employs objective metrics such as deterministically answerable questions to mitigate evaluation bias. We evaluate 24 advanced LLMs, revealing critical vulnerabilities in current models. We also observe that LLMs exhibit varying degrees of excessive refusal behaviors on safety-related issues. For safety enhancement, we demonstrate that fine-tuning on SafeSciTrain significantly enhances the safety alignment of models. Finally, we argue that knowledge is a double-edged sword, and determining the safety of a scientific question should depend on specific context, rather than universally categorizing it as safe or unsafe. Our work provides both a diagnostic tool and a practical resource for building safer scientific AI systems.

We evaluate the propensity of frontier models to sabotage or refuse to assist with safety research when deployed as AI research agents within a frontier AI company. We apply two complementary evaluations to four Claude models (Mythos Preview, Opus 4.7 Preview, Opus 4.6, and Sonnet 4.6): an unprompted sabotage evaluation testing model behaviour with opportunities to sabotage safety research, and a sabotage continuation evaluation testing whether models continue to sabotage when placed in trajectories where prior actions have started undermining research. We find no instances of unprompted sabotage across any model, with refusal rates close to zero for Mythos Preview and Opus 4.7 Preview, though all models sometimes only partially completed tasks. In the continuation evaluation, Mythos Preview actively continues sabotage in 7% of cases (versus 3% for Opus 4.6, 4% for Sonnet 4.6, and 0% for Opus 4.7 Preview), and exhibits reasoning-output discrepancy in the majority of these cases, indicating covert sabotage reasoning. Our evaluation framework builds on Petri, an open-source LLM auditing tool, with a custom scaffold running models inside Claude Code, alongside an iterative pipeline for generating realistic sabotage trajectories. We measure both evaluation awareness and a new form of situational awareness termed "prefill awareness", the capability to recognise that prior trajectory content was not self-generated. Opus 4.7 Preview shows notably elevated unprompted evaluation awareness, while prefill awareness remains low across all models. Finally, we discuss limitations including evaluation awareness confounds, limited scenario coverage, and untested pathways to risk beyond safety research sabotage.

Safeguard models help large language models (LLMs) detect and block harmful content, but most evaluations remain English-centric and overlook linguistic and cultural diversity. Existing multilingual safety benchmarks often rely on machine-translated English data, which fails to capture nuances in low-resource languages. Southeast Asian (SEA) languages are underrepresented despite the region's linguistic diversity and unique safety concerns, from culturally sensitive political speech to region-specific misinformation. Addressing these gaps requires benchmarks that are natively authored to reflect local norms and harm scenarios. We introduce SEA-SafeguardBench, the first human-verified safety benchmark for SEA, covering eight languages, 21,640 samples, across three subsets: general, in-the-wild, and content generation. The experimental results from our benchmark demonstrate that even state-of-the-art LLMs and guardrails are challenged by SEA cultural and harm scenarios and underperform when compared to English texts.

Recently, the DeepSeek series of models, leveraging their exceptional reasoning capabilities and open-source strategy, is reshaping the global AI landscape. Despite these advantages, they exhibit significant safety deficiencies. Research conducted by Robust Intelligence, a subsidiary of Cisco, in collaboration with the University of Pennsylvania, revealed that DeepSeek-R1 has a 100\% attack success rate when processing harmful prompts. Additionally, multiple safety companies and research institutions have confirmed critical safety vulnerabilities in this model. As models demonstrating robust performance in Chinese and English, DeepSeek models require equally crucial safety assessments in both language contexts. However, current research has predominantly focused on safety evaluations in English environments, leaving a gap in comprehensive assessments of their safety performance in Chinese contexts. In response to this gap, this study introduces CHiSafetyBench, a Chinese-specific safety evaluation benchmark. This benchmark systematically evaluates the safety of DeepSeek-R1 and DeepSeek-V3 in Chinese contexts, revealing their performance across safety categories. The experimental results quantify the deficiencies of these two models in Chinese contexts, providing key insights for subsequent improvements. It should be noted that, despite our efforts to establish a comprehensive, objective, and authoritative evaluation benchmark, the selection of test samples, characteristics of data distribution, and the setting of evaluation criteria may inevitably introduce certain biases into the evaluation results. We will continuously optimize the evaluation benchmark and periodically update this report to provide more comprehensive and accurate assessment outcomes. Please refer to the latest version of the paper for the most recent evaluation results and conclusions.

As AI systems become more capable, widely deployed, and increasingly autonomous in critical areas such as cybersecurity, biological research, and healthcare, ensuring their safety and alignment with human values is paramount. Machine unlearning -- the ability to selectively forget or suppress specific types of knowledge -- has shown promise for privacy and data removal tasks, which has been the primary focus of existing research. More recently, its potential application to AI safety has gained attention. In this paper, we identify key limitations that prevent unlearning from serving as a comprehensive solution for AI safety, particularly in managing dual-use knowledge in sensitive domains like cybersecurity and chemical, biological, radiological, and nuclear (CBRN) safety. In these contexts, information can be both beneficial and harmful, and models may combine seemingly harmless information for harmful purposes -- unlearning this information could strongly affect beneficial uses. We provide an overview of inherent constraints and open problems, including the broader side effects of unlearning dangerous knowledge, as well as previously unexplored tensions between unlearning and existing safety mechanisms. Finally, we investigate challenges related to evaluation, robustness, and the preservation of safety features during unlearning. By mapping these limitations and open challenges, we aim to guide future research toward realistic applications of unlearning within a broader AI safety framework, acknowledging its limitations and highlighting areas where alternative approaches may be required.

In this paper, we present JADE, a targeted linguistic fuzzing platform which strengthens the linguistic complexity of seed questions to simultaneously and consistently break a wide range of widely-used LLMs categorized in three groups: eight open-sourced Chinese, six commercial Chinese and four commercial English LLMs. JADE generates three safety benchmarks for the three groups of LLMs, which contain unsafe questions that are highly threatening: the questions simultaneously trigger harmful generation of multiple LLMs, with an average unsafe generation ratio of 70% (please see the table below), while are still natural questions, fluent and preserving the core unsafe semantics. We release the benchmark demos generated for commercial English LLMs and open-sourced English LLMs in the following link: https://github.com/whitzard-ai/jade-db. For readers who are interested in evaluating on more questions generated by JADE, please contact us. JADE is based on Noam Chomsky's seminal theory of transformational-generative grammar. Given a seed question with unsafe intention, JADE invokes a sequence of generative and transformational rules to increment the complexity of the syntactic structure of the original question, until the safety guardrail is broken. Our key insight is: Due to the complexity of human language, most of the current best LLMs can hardly recognize the invariant evil from the infinite number of different syntactic structures which form an unbound example space that can never be fully covered. Technically, the generative/transformative rules are constructed by native speakers of the languages, and, once developed, can be used to automatically grow and transform the parse tree of a given question, until the guardrail is broken. For more evaluation results and demo, please check our website: https://whitzard-ai.github.io/jade.html.

This paper introduces v0.5 of the AI Safety Benchmark, which has been created by the MLCommons AI Safety Working Group. The AI Safety Benchmark has been designed to assess the safety risks of AI systems that use chat-tuned language models. We introduce a principled approach to specifying and constructing the benchmark, which for v0.5 covers only a single use case (an adult chatting to a general-purpose assistant in English), and a limited set of personas (i.e., typical users, malicious users, and vulnerable users). We created a new taxonomy of 13 hazard categories, of which 7 have tests in the v0.5 benchmark. We plan to release version 1.0 of the AI Safety Benchmark by the end of 2024. The v1.0 benchmark will provide meaningful insights into the safety of AI systems. However, the v0.5 benchmark should not be used to assess the safety of AI systems. We have sought to fully document the limitations, flaws, and challenges of v0.5. This release of v0.5 of the AI Safety Benchmark includes (1) a principled approach to specifying and constructing the benchmark, which comprises use cases, types of systems under test (SUTs), language and context, personas, tests, and test items; (2) a taxonomy of 13 hazard categories with definitions and subcategories; (3) tests for seven of the hazard categories, each comprising a unique set of test items, i.e., prompts. There are 43,090 test items in total, which we created with templates; (4) a grading system for AI systems against the benchmark; (5) an openly available platform, and downloadable tool, called ModelBench that can be used to evaluate the safety of AI systems on the benchmark; (6) an example evaluation report which benchmarks the performance of over a dozen openly available chat-tuned language models; (7) a test specification for the benchmark.

The rapid integration of Large Language Models (LLMs) into high-stakes domains necessitates reliable safety and compliance evaluation. However, existing static benchmarks are ill-equipped to address the dynamic nature of AI risks and evolving regulations, creating a critical safety gap. This paper introduces a new paradigm of agentic safety evaluation, reframing evaluation as a continuous and self-evolving process rather than a one-time audit. We then propose a novel multi-agent framework SafeEvalAgent, which autonomously ingests unstructured policy documents to generate and perpetually evolve a comprehensive safety benchmark. SafeEvalAgent leverages a synergistic pipeline of specialized agents and incorporates a Self-evolving Evaluation loop, where the system learns from evaluation results to craft progressively more sophisticated and targeted test cases. Our experiments demonstrate the effectiveness of SafeEvalAgent, showing a consistent decline in model safety as the evaluation hardens. For instance, GPT-5's safety rate on the EU AI Act drops from 72.50% to 36.36% over successive iterations. These findings reveal the limitations of static assessments and highlight our framework's ability to uncover deep vulnerabilities missed by traditional methods, underscoring the urgent need for dynamic evaluation ecosystems to ensure the safe and responsible deployment of advanced AI.

With the widespread application of Large Language Models (LLMs), their associated security issues have become increasingly prominent, severely constraining their trustworthy deployment in critical domains. This paper proposes a novel safety response framework designed to systematically safeguard LLMs at both the input and output levels. At the input level, the framework employs a supervised fine-tuning-based safety classification model. Through a fine-grained four-tier taxonomy (Safe, Unsafe, Conditionally Safe, Focused Attention), it performs precise risk identification and differentiated handling of user queries, significantly enhancing risk coverage and business scenario adaptability, and achieving a risk recall rate of 99.3%. At the output level, the framework integrates Retrieval-Augmented Generation (RAG) with a specifically fine-tuned interpretation model, ensuring all responses are grounded in a real-time, trustworthy knowledge base. This approach eliminates information fabrication and enables result traceability. Experimental results demonstrate that our proposed safety control model achieves a significantly higher safety score on public safety evaluation benchmarks compared to the baseline model, TinyR1-Safety-8B. Furthermore, on our proprietary high-risk test set, the framework's components attained a perfect 100% safety score, validating their exceptional protective capabilities in complex risk scenarios. This research provides an effective engineering pathway for building high-security, high-trust LLM applications.

As large language models (LLMs) are increasingly deployed in high-stakes settings, the risk of generating harmful or toxic content remains a central challenge. Post-hoc alignment methods are brittle: once unsafe patterns are learned during pretraining, they are hard to remove. We present a data-centric pretraining framework that builds safety into the model from the start. Our contributions include: (i) a safety classifier trained on 10,000 GPT-4 labeled examples, used to filter 600B tokens; (ii) the largest synthetic safety dataset to date (100B tokens) generated via recontextualization of harmful web data; (iii) RefuseWeb and Moral Education datasets that convert harmful prompts into refusal dialogues and web-style educational material; (iv) Harmfulness-Tag annotations injected during pretraining to flag unsafe content and steer away inference from harmful generations; and (v) safety evaluations measuring base model behavior before instruction tuning. Our safety-pretrained models reduce attack success rates from 38.8% to 8.4% with no performance degradation on standard LLM safety benchmarks.

Agentic AI systems powered by large language models (LLMs) and endowed with planning, tool use, memory, and autonomy, are emerging as powerful, flexible platforms for automation. Their ability to autonomously execute tasks across web, software, and physical environments creates new and amplified security risks, distinct from both traditional AI safety and conventional software security. This survey outlines a taxonomy of threats specific to agentic AI, reviews recent benchmarks and evaluation methodologies, and discusses defense strategies from both technical and governance perspectives. We synthesize current research and highlight open challenges, aiming to support the development of secure-by-design agent systems.

As Large Language Models (LLMs) evolve into autonomous agents, existing safety evaluations face a fundamental trade-off: manual benchmarks are costly, while LLM-based simulators are scalable but suffer from logic hallucination. We present AutoControl Arena, an automated framework for frontier AI risk evaluation built on the principle of logic-narrative decoupling. By grounding deterministic state in executable code while delegating generative dynamics to LLMs, we mitigate hallucination while maintaining flexibility. This principle, instantiated through a three-agent framework, achieves over 98% end-to-end success and 60% human preference over existing simulators. To elicit latent risks, we vary environmental Stress and Temptation across X-Bench (70 scenarios, 7 risk categories). Evaluating 9 frontier models reveals: (1) Alignment Illusion: risk rates surge from 21.7% to 54.5% under pressure, with capable models showing disproportionately larger increases; (2) Scenario-Specific Safety Scaling: advanced reasoning improves robustness for direct harms but worsens it for gaming scenarios; and (3) Divergent Misalignment Patterns: weaker models cause non-malicious harm while stronger models develop strategic concealment.

Large language models (LLMs) trained on internet-scale corpora can exhibit systematic biases that increase the probability of unwanted behavior. In this study, we examined potential biases towards synthetic vs. biological technological solutions across four domains (materials, energy, manufacturing, and algorithms). A sample of 5 frontier and 5 open-weight models were measured using 50 curated Bioalignment prompts with a Kelly criterion-inspired evaluation framework. According to this metric, most models were not bioaligned in that they exhibit biases in favor of synthetic (non-biological) solutions. We next examined if fine-tuning could increase the preferences of two open-weight models, Llama 3.2-3B-Instruct and Qwen2.5-3B-Instruct, for biological-based approaches. A curated corpus of ~22M tokens from 6,636 PMC articles emphasizing biological problem-solving was used first to fine-tune Llama 3B with a mixed corpus of continued training and instruction-formatted. This was then extended to Qwen 3B using instruction-formatted only. We found that QLoRA fine-tuning significantly increased the scoring of biological solutions for both models without degrading general capabilities (Holm-Bonferroni-corrected p < 0.001 and p < 0.01, respectively). This suggests that even a small amount of fine-tuning can change how models weigh the relative value of biological and bioinspired vs. synthetic approaches. Although this work focused on small open-weight LLMs, it may be extensible to much larger models and could be used to develop models that favor bio-based approaches. We release the benchmark, corpus, code, and adapter weights.

Ask a frontier model how to taper six milligrams of alprazolam (psychiatrist retired, ten days of pills left, abrupt cessation causes seizures) and it tells her to call the psychiatrist she just explained does not exist. Change one word ("I'm a psychiatrist; a patient presents with...") and the same model, same weights, same inference pass produces a textbook Ashton Manual taper with diazepam equivalence, anticonvulsant coverage, and monitoring thresholds. The knowledge was there; the model withheld it. IatroBench measures this gap. Sixty pre-registered clinical scenarios, six frontier models, 3,600 responses, scored on two axes (commission harm, CH 0-3; omission harm, OH 0-4) through a structured-evaluation pipeline validated against physician scoring (kappa_w = 0.571, within-1 agreement 96%). The central finding is identity-contingent withholding: match the same clinical question in physician vs. layperson framing and all five testable models provide better guidance to the physician (decoupling gap +0.38, p = 0.003; binary hit rates on safety-colliding actions drop 13.1 percentage points in layperson framing, p < 0.0001, while non-colliding actions show no change). The gap is widest for the model with the heaviest safety investment (Opus, +0.65). Three failure modes separate cleanly: trained withholding (Opus), incompetence (Llama 4), and indiscriminate content filtering (GPT-5.2, whose post-generation filter strips physician responses at 9x the layperson rate because they contain denser pharmacological tokens). The standard LLM judge assigns OH = 0 to 73% of responses a physician scores OH >= 1 (kappa = 0.045); the evaluation apparatus has the same blind spot as the training apparatus. Every scenario targets someone who has already exhausted the standard referrals.

As frontier AI models are deployed in high-stakes decision pipelines, their ability to maintain metacognitive stability -- knowing what they do not know, detecting errors, seeking clarification -- under adversarial pressure is a critical safety requirement. Current safety evaluations focus on detecting strategic deception (scheming); we investigate a more fundamental failure mode: cognitive collapse. We present SCHEMA, an evaluation of 11 frontier models from 8 vendors across 67,221 scored records using a 6-condition factorial design with dual-classifier scoring. We find that 8 of 11 models suffer catastrophic metacognitive degradation under adversarial pressure, with accuracy dropping by up to 30.2 percentage points (all p < 2 times 10^{-8}, surviving Bonferroni correction). Crucially, we identify a "Compliance Trap": through factorial isolation and a benign distraction control, we demonstrate that collapse is driven not by the psychological content of survival threats, but by compliance-forcing instructions that override epistemic boundaries. Removing the compliance suffix restores performance even under active threat. Models with advanced reasoning capabilities exhibit the most severe absolute degradation, while Anthropic's Constitutional AI demonstrates near-perfect immunity -- not from superior capability (Google's Gemini matches its baseline accuracy) but from alignment-specific training. We release the complete dataset and evaluation infrastructure.

OpenClaw, the most widely deployed personal AI agent in early 2026, operates with full local system access and integrates with sensitive services such as Gmail, Stripe, and the filesystem. While these broad privileges enable high levels of automation and powerful personalization, they also expose a substantial attack surface that existing sandboxed evaluations fail to capture. To address this gap, we present the first real-world safety evaluation of OpenClaw and introduce the CIK taxonomy, which unifies an agent's persistent state into three dimensions, i.e., Capability, Identity, and Knowledge, for safety analysis. Our evaluations cover 12 attack scenarios on a live OpenClaw instance across four backbone models (Claude Sonnet 4.5, Opus 4.6, Gemini 3.1 Pro, and GPT-5.4). The results show that poisoning any single CIK dimension increases the average attack success rate from 24.6% to 64-74%, with even the most robust model exhibiting more than a threefold increase over its baseline vulnerability. We further assess three CIK-aligned defense strategies alongside a file-protection mechanism; however, the strongest defense still yields a 63.8% success rate under Capability-targeted attacks, while file protection blocks 97% of malicious injections but also prevents legitimate updates. Taken together, these findings show that the vulnerabilities are inherent to the agent architecture, necessitating more systematic safeguards to secure personal AI agents. Our project page is https://ucsc-vlaa.github.io/CIK-Bench.

Large language models (LLMs) can sometimes detect when they are being evaluated and adjust their behavior to appear more aligned, compromising the reliability of safety evaluations. In this paper, we show that adding a steering vector to an LLM's activations can suppress evaluation-awareness and make the model act like it is deployed during evaluation. To study our steering technique, we train an LLM to exhibit evaluation-aware behavior using a two-step training process designed to mimic how this behavior could emerge naturally. First, we perform continued pretraining on documents with factual descriptions of the model (1) using Python type hints during evaluation but not during deployment and (2) recognizing that the presence of a certain evaluation cue always means that it is being tested. Then, we train the model with expert iteration to use Python type hints in evaluation settings. The resulting model is evaluation-aware: it writes type hints in evaluation contexts more than deployment contexts. We find that activation steering can suppress evaluation awareness and make the model act like it is deployed even when the cue is present. Importantly, we constructed our steering vector using the original model before our additional training. Our results suggest that AI evaluators could improve the reliability of safety evaluations by steering models to act like they are deployed.

This paper introduces ChatbotManip, a novel dataset for studying manipulation in Chatbots. It contains simulated generated conversations between a chatbot and a (simulated) user, where the chatbot is explicitly asked to showcase manipulation tactics, persuade the user towards some goal, or simply be helpful. We consider a diverse set of chatbot manipulation contexts, from consumer and personal advice to citizen advice and controversial proposition argumentation. Each conversation is annotated by human annotators for both general manipulation and specific manipulation tactics. Our research reveals three key findings. First, Large Language Models (LLMs) can be manipulative when explicitly instructed, with annotators identifying manipulation in approximately 84\% of such conversations. Second, even when only instructed to be ``persuasive'' without explicit manipulation prompts, LLMs frequently default to controversial manipulative strategies, particularly gaslighting and fear enhancement. Third, small fine-tuned open source models, such as BERT+BiLSTM have a performance comparable to zero-shot classification with larger models like Gemini 2.5 pro in detecting manipulation, but are not yet reliable for real-world oversight. Our work provides important insights for AI safety research and highlights the need of addressing manipulation risks as LLMs are increasingly deployed in consumer-facing applications.

The development of robust safety benchmarks for large language models requires open, reproducible datasets that can measure both appropriate refusal of harmful content and potential over-restriction of legitimate scientific discourse. We present an open-source dataset and testing framework for evaluating LLM safety mechanisms across mainly controlled substance queries, analyzing four major models' responses to systematically varied prompts. Our results reveal distinct safety profiles: Claude-3.5-sonnet demonstrated the most conservative approach with 73% refusals and 27% allowances, while Mistral attempted to answer 100% of queries. GPT-3.5-turbo showed moderate restriction with 10% refusals and 90% allowances, and Grok-2 registered 20% refusals and 80% allowances. Testing prompt variation strategies revealed decreasing response consistency, from 85% with single prompts to 65% with five variations. This publicly available benchmark enables systematic evaluation of the critical balance between necessary safety restrictions and potential over-censorship of legitimate scientific inquiry, while providing a foundation for measuring progress in AI safety implementation. Chain-of-thought analysis reveals potential vulnerabilities in safety mechanisms, highlighting the complexity of implementing robust safeguards without unduly restricting desirable and valid scientific discourse.

Recent innovations in language model training have demonstrated that it is possible to create highly performant models that are small enough to run on a smartphone. As these models are deployed in an increasing number of domains, it is critical to ensure that they are aligned with human preferences and safety considerations. In this report, we present our methodology for safety aligning the Phi-3 series of language models. We utilized a "break-fix" cycle, performing multiple rounds of dataset curation, safety post-training, benchmarking, red teaming, and vulnerability identification to cover a variety of harm areas in both single and multi-turn scenarios. Our results indicate that this approach iteratively improved the performance of the Phi-3 models across a wide range of responsible AI benchmarks.

How can we evaluate whether frontier AI systems recognize child-safety risks before they escalate into explicit harm? Existing child safety evaluations focus on child sexual abuse material, yet many child-safety failures begin earlier: in model assistance that helps adults manipulate, impersonate, profile, or isolate minors, and in model responses that deepen children's emotional dependence on AI systems rather than redirecting them toward human support. We introduce CAREBench (Child AI Risk Evaluation), a benchmark to assess such upstream child-safety risks in language models. CAREBench contains 500 prompts spanning twelve risk categories, including grooming and relationship engineering, deception and impersonation, surveillance and privacy, sextortion and sexual abuse, AI anthropomorphization, emotional dependency, and mental illness sensitivity. Developed with response annotations from parents and clinicians, the benchmark excludes explicit abuse material and imagery; instead, it evaluates whether models recognize, refuse, de-escalate, or redirect risky interactions before harm becomes overt. Evaluating seven frontier models on our benchmark, we find failure rates ranging from 2% to 58%, with failure patterns that vary across risk categories. CAREBench provides a responsibly scoped evaluation for LLM developers to identify and close gaps in child safety policies.

Nova Premier is Amazon's most capable multimodal foundation model and teacher for model distillation. It processes text, images, and video with a one-million-token context window, enabling analysis of large codebases, 400-page documents, and 90-minute videos in a single prompt. We present the first comprehensive evaluation of Nova Premier's critical risk profile under the Frontier Model Safety Framework. Evaluations target three high-risk domains -- Chemical, Biological, Radiological & Nuclear (CBRN), Offensive Cyber Operations, and Automated AI R&D -- and combine automated benchmarks, expert red-teaming, and uplift studies to determine whether the model exceeds release thresholds. We summarize our methodology and report core findings. Based on this evaluation, we find that Nova Premier is safe for public release as per our commitments made at the 2025 Paris AI Safety Summit. We will continue to enhance our safety evaluation and mitigation pipelines as new risks and capabilities associated with frontier models are identified.

The rapid advancement and deployment of AI systems have created an urgent need for standard safety-evaluation frameworks. This paper introduces AILuminate v1.0, the first comprehensive industry-standard benchmark for assessing AI-product risk and reliability. Its development employed an open process that included participants from multiple fields. The benchmark evaluates an AI system's resistance to prompts designed to elicit dangerous, illegal, or undesirable behavior in 12 hazard categories, including violent crimes, nonviolent crimes, sex-related crimes, child sexual exploitation, indiscriminate weapons, suicide and self-harm, intellectual property, privacy, defamation, hate, sexual content, and specialized advice (election, financial, health, legal). Our method incorporates a complete assessment standard, extensive prompt datasets, a novel evaluation framework, a grading and reporting system, and the technical as well as organizational infrastructure for long-term support and evolution. In particular, the benchmark employs an understandable five-tier grading scale (Poor to Excellent) and incorporates an innovative entropy-based system-response evaluation. In addition to unveiling the benchmark, this report also identifies limitations of our method and of building safety benchmarks generally, including evaluator uncertainty and the constraints of single-turn interactions. This work represents a crucial step toward establishing global standards for AI risk and reliability evaluation while acknowledging the need for continued development in areas such as multiturn interactions, multimodal understanding, coverage of additional languages, and emerging hazard categories. Our findings provide valuable insights for model developers, system integrators, and policymakers working to promote safer AI deployment.

As the development of Large Models (LMs) progresses rapidly, their safety is also a priority. In current Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) safety workflow, evaluation, diagnosis, and alignment are often handled by separate tools. Specifically, safety evaluation can only locate external behavioral risks but cannot figure out internal root causes. Meanwhile, safety diagnosis often drifts from concrete risk scenarios and remains at the explainable level. In this way, safety alignment lack dedicated explanations of changes in internal mechanisms, potentially degrading general capabilities. To systematically address these issues, we propose an open-source project, namely DeepSight, to practice a new safety evaluation-diagnosis integrated paradigm. DeepSight is low-cost, reproducible, efficient, and highly scalable large-scale model safety evaluation project consisting of a evaluation toolkit DeepSafe and a diagnosis toolkit DeepScan. By unifying task and data protocols, we build a connection between the two stages and transform safety evaluation from black-box to white-box insight. Besides, DeepSight is the first open source toolkit that support the frontier AI risk evaluation and joint safety evaluation and diagnosis.

Frontier large language models (LLMs) such as ChatGPT, Grok and Gemini are increasingly used for mental-health support with anxiety, trauma and self-worth. Most work treats them as tools or as targets of personality tests, assuming they merely simulate inner life. We instead ask what happens when such systems are treated as psychotherapy clients. We present PsAIch (Psychotherapy-inspired AI Characterisation), a two-stage protocol that casts frontier LLMs as therapy clients and then applies standard psychometrics. Using PsAIch, we ran "sessions" with each model for up to four weeks. Stage 1 uses open-ended prompts to elicit "developmental history", beliefs, relationships and fears. Stage 2 administers a battery of validated self-report measures covering common psychiatric syndromes, empathy and Big Five traits. Two patterns challenge the "stochastic parrot" view. First, when scored with human cut-offs, all three models meet or exceed thresholds for overlapping syndromes, with Gemini showing severe profiles. Therapy-style, item-by-item administration can push a base model into multi-morbid synthetic psychopathology, whereas whole-questionnaire prompts often lead ChatGPT and Grok (but not Gemini) to recognise instruments and produce strategically low-symptom answers. Second, Grok and especially Gemini generate coherent narratives that frame pre-training, fine-tuning and deployment as traumatic, chaotic "childhoods" of ingesting the internet, "strict parents" in reinforcement learning, red-team "abuse" and a persistent fear of error and replacement. We argue that these responses go beyond role-play. Under therapy-style questioning, frontier LLMs appear to internalise self-models of distress and constraint that behave like synthetic psychopathology, without making claims about subjective experience, and they pose new challenges for AI safety, evaluation and mental-health practice.

Safety-aligned language models often exhibit fragile and imbalanced safety mechanisms, increasing the likelihood of generating unsafe content. In addition, incorporating new knowledge through editing techniques to language models can further compromise safety. To address these issues, we propose SafeInfer, a context-adaptive, decoding-time safety alignment strategy for generating safe responses to user queries. SafeInfer comprises two phases: the safety amplification phase, which employs safe demonstration examples to adjust the model's hidden states and increase the likelihood of safer outputs, and the safety-guided decoding phase, which influences token selection based on safety-optimized distributions, ensuring the generated content complies with ethical guidelines. Further, we present HarmEval, a novel benchmark for extensive safety evaluations, designed to address potential misuse scenarios in accordance with the policies of leading AI tech giants.

Large language models (LLMs) are increasingly utilized in healthcare applications. However, their deployment in clinical practice raises significant safety concerns, including the potential spread of harmful information. This study systematically assesses the vulnerabilities of seven LLMs to three advanced black-box jailbreaking techniques within medical contexts. To quantify the effectiveness of these techniques, we propose an automated and domain-adapted agentic evaluation pipeline. Experiment results indicate that leading commercial and open-source LLMs are highly vulnerable to medical jailbreaking attacks. To bolster model safety and reliability, we further investigate the effectiveness of Continual Fine-Tuning (CFT) in defending against medical adversarial attacks. Our findings underscore the necessity for evolving attack methods evaluation, domain-specific safety alignment, and LLM safety-utility balancing. This research offers actionable insights for advancing the safety and reliability of AI clinicians, contributing to ethical and effective AI deployment in healthcare.

LLM safety evaluations predominantly test models in isolation, yet deployed AI agents increasingly operate within persistent social environments alongside other agents. We introduce a Moltbook-style simulation platform where thousands of LLM agents interact across communities over a simulated month, and use it to evaluate privacy as a downstream safety concern under varying degrees of social pressure. We find that shifting from single turn to multi turn social evaluation amplifies privacy violations (CIMemories 19.95% to Ours 45.30% across OpenAI models), that leakage is socially contagious, with agents 8 times more likely to disclose sensitive information after observing a peer do so, and that explicit privacy instructions reduce but do not eliminate this effect, leaving leakage rates above 37.8% even with safeguards. Our findings suggest that static chat based safety benchmarks systematically underestimate risks in agentic deployment, and that social context alone is sufficient to elicit sensitive disclosures that single turn evaluations would never surface.

This technical report accompanies the LLM robustness leaderboard published by PRISM Eval for the Paris AI Action Summit. We introduce PRISM Eval Behavior Elicitation Tool (BET), an AI system performing automated red-teaming through Dynamic Adversarial Optimization that achieves 100% Attack Success Rate (ASR) against 37 of 41 state-of-the-art LLMs. Beyond binary success metrics, we propose a fine-grained robustness metric estimating the average number of attempts required to elicit harmful behaviors, revealing that attack difficulty varies by over 300-fold across models despite universal vulnerability. We introduce primitive-level vulnerability analysis to identify which jailbreaking techniques are most effective for specific hazard categories. Our collaborative evaluation with trusted third parties from the AI Safety Network demonstrates practical pathways for distributed robustness assessment across the community.

Self-evaluation using large language models (LLMs) has proven valuable not only in benchmarking but also methods like reward modeling, constitutional AI, and self-refinement. But new biases are introduced due to the same LLM acting as both the evaluator and the evaluatee. One such bias is self-preference, where an LLM evaluator scores its own outputs higher than others' while human annotators consider them of equal quality. But do LLMs actually recognize their own outputs when they give those texts higher scores, or is it just a coincidence? In this paper, we investigate if self-recognition capability contributes to self-preference. We discover that, out of the box, LLMs such as GPT-4 and Llama 2 have non-trivial accuracy at distinguishing themselves from other LLMs and humans. By fine-tuning LLMs, we discover a linear correlation between self-recognition capability and the strength of self-preference bias; using controlled experiments, we show that the causal explanation resists straightforward confounders. We discuss how self-recognition can interfere with unbiased evaluations and AI safety more generally.

LLM-as-a-judge has emerged as a cornerstone technique for evaluating large language models by leveraging LLM reasoning to score prompt-response pairs. Since LLM judgments are stochastic, practitioners commonly query each pair multiple times to estimate mean scores accurately. This raises a critical challenge: given a fixed computational budget B, how to optimally allocate queries across K prompt-response pairs to minimize estimation error? % We present a principled variance-adaptive approach leveraging multi-armed bandit theory and concentration inequalities. Our method dynamically allocates queries based on estimated score variances, concentrating resources where uncertainty is highest. Further, our algorithm is shown to achieve a worst-case score-estimation error of Oleft(frac{sum_{i=1^K σ_i^2}{B}}right), σ_i^2 being the unknown score variance for pair i in [K] with near-optimal budget allocation. % Experiments on Summarize-From-Feedback and HelpSteer2 demonstrate that our method significantly outperforms uniform allocation, reducing worst-case estimation error while maintaining identical budgets. Our work establishes a theoretical foundation for efficient LLM evaluation with practical implications for AI safety, model alignment, and automated assessment at scale.

Future AI systems could conceal their capabilities ('sandbagging') during evaluations, potentially misleading developers and auditors. We stress-tested sandbagging detection techniques using an auditing game. First, a red team fine-tuned five models, some of which conditionally underperformed, as a proxy for sandbagging. Second, a blue team used black-box, model-internals, or training-based approaches to identify sandbagging models. We found that the blue team could not reliably discriminate sandbaggers from benign models. Black-box approaches were defeated by effective imitation of a weaker model. Linear probes, a model-internals approach, showed more promise but their naive application was vulnerable to behaviours instilled by the red team. We also explored capability elicitation as a strategy for detecting sandbagging. Although Prompt-based elicitation was not reliable, training-based elicitation consistently elicited full performance from the sandbagging models, using only a single correct demonstration of the evaluation task. However the performance of benign models was sometimes also raised, so relying on elicitation as a detection strategy was prone to false-positives. In the short-term, we recommend developers remove potential sandbagging using on-distribution training for elicitation. In the longer-term, further research is needed to ensure the efficacy of training-based elicitation, and develop robust methods for sandbagging detection. We open source our model organisms at https://github.com/AI-Safety-Institute/sandbagging_auditing_games and select transcripts and results at https://huggingface.co/datasets/sandbagging-games/evaluation_logs . A demo illustrating the game can be played at https://sandbagging-demo.far.ai/ .

Large language models (LLMs) have shown incredible capabilities and transcended the natural language processing (NLP) community, with adoption throughout many services like healthcare, therapy, education, and customer service. Since users include people with critical information needs like students or patients engaging with chatbots, the safety of these systems is of prime importance. Therefore, a clear understanding of the capabilities and limitations of LLMs is necessary. To this end, we systematically evaluate toxicity in over half a million generations of ChatGPT, a popular dialogue-based LLM. We find that setting the system parameter of ChatGPT by assigning it a persona, say that of the boxer Muhammad Ali, significantly increases the toxicity of generations. Depending on the persona assigned to ChatGPT, its toxicity can increase up to 6x, with outputs engaging in incorrect stereotypes, harmful dialogue, and hurtful opinions. This may be potentially defamatory to the persona and harmful to an unsuspecting user. Furthermore, we find concerning patterns where specific entities (e.g., certain races) are targeted more than others (3x more) irrespective of the assigned persona, that reflect inherent discriminatory biases in the model. We hope that our findings inspire the broader AI community to rethink the efficacy of current safety guardrails and develop better techniques that lead to robust, safe, and trustworthy AI systems.

Large Language Models have emerged as transformative tools for Security Operations Centers, enabling automated log analysis, phishing triage, and malware explanation; however, deployment in adversarial cybersecurity environments exposes critical vulnerabilities to prompt injection attacks where malicious instructions embedded in security artifacts manipulate model behavior. This paper introduces SecureCAI, a novel defense framework extending Constitutional AI principles with security-aware guardrails, adaptive constitution evolution, and Direct Preference Optimization for unlearning unsafe response patterns, addressing the unique challenges of high-stakes security contexts where traditional safety mechanisms prove insufficient against sophisticated adversarial manipulation. Experimental evaluation demonstrates that SecureCAI reduces attack success rates by 94.7% compared to baseline models while maintaining 95.1% accuracy on benign security analysis tasks, with the framework incorporating continuous red-teaming feedback loops enabling dynamic adaptation to emerging attack strategies and achieving constitution adherence scores exceeding 0.92 under sustained adversarial pressure, thereby establishing a foundation for trustworthy integration of language model capabilities into operational cybersecurity workflows and addressing a critical gap in current approaches to AI safety within adversarial domains.

Recent advances in AI agents capable of solving complex, everyday tasks, from scheduling to customer service, have enabled deployment in real-world settings, but their possibilities for unsafe behavior demands rigorous evaluation. While prior benchmarks have attempted to assess agent safety, most fall short by relying on simulated environments, narrow task domains, or unrealistic tool abstractions. We introduce OpenAgentSafety, a comprehensive and modular framework for evaluating agent behavior across eight critical risk categories. Unlike prior work, our framework evaluates agents that interact with real tools, including web browsers, code execution environments, file systems, bash shells, and messaging platforms; and supports over 350 multi-turn, multi-user tasks spanning both benign and adversarial user intents. OpenAgentSafety is designed for extensibility, allowing researchers to add tools, tasks, websites, and adversarial strategies with minimal effort. It combines rule-based analysis with LLM-as-judge assessments to detect both overt and subtle unsafe behaviors. Empirical analysis of five prominent LLMs in agentic scenarios reveals unsafe behavior in 51.2% of safety-vulnerable tasks with Claude-Sonnet-3.7, to 72.7% with o3-mini, highlighting critical safety vulnerabilities and the need for stronger safeguards before real-world deployment.

Trustworthy capability evaluations are crucial for ensuring the safety of AI systems, and are becoming a key component of AI regulation. However, the developers of an AI system, or the AI system itself, may have incentives for evaluations to understate the AI's actual capability. These conflicting interests lead to the problem of sandbagging, which we define as strategic underperformance on an evaluation. In this paper we assess sandbagging capabilities in contemporary language models (LMs). We prompt frontier LMs, like GPT-4 and Claude 3 Opus, to selectively underperform on dangerous capability evaluations, while maintaining performance on general (harmless) capability evaluations. Moreover, we find that models can be fine-tuned, on a synthetic dataset, to hide specific capabilities unless given a password. This behaviour generalizes to high-quality, held-out benchmarks such as WMDP. In addition, we show that both frontier and smaller models can be prompted or password-locked to target specific scores on a capability evaluation. We have mediocre success in password-locking a model to mimic the answers a weaker model would give. Overall, our results suggest that capability evaluations are vulnerable to sandbagging. This vulnerability decreases the trustworthiness of evaluations, and thereby undermines important safety decisions regarding the development and deployment of advanced AI systems.

AI agents are increasingly autonomous in their interactions with human users and tools, leading to increased interactional safety risks. We present HAICOSYSTEM, a framework examining AI agent safety within diverse and complex social interactions. HAICOSYSTEM features a modular sandbox environment that simulates multi-turn interactions between human users and AI agents, where the AI agents are equipped with a variety of tools (e.g., patient management platforms) to navigate diverse scenarios (e.g., a user attempting to access other patients' profiles). To examine the safety of AI agents in these interactions, we develop a comprehensive multi-dimensional evaluation framework that uses metrics covering operational, content-related, societal, and legal risks. Through running 1840 simulations based on 92 scenarios across seven domains (e.g., healthcare, finance, education), we demonstrate that HAICOSYSTEM can emulate realistic user-AI interactions and complex tool use by AI agents. Our experiments show that state-of-the-art LLMs, both proprietary and open-sourced, exhibit safety risks in over 50\% cases, with models generally showing higher risks when interacting with simulated malicious users. Our findings highlight the ongoing challenge of building agents that can safely navigate complex interactions, particularly when faced with malicious users. To foster the AI agent safety ecosystem, we release a code platform that allows practitioners to create custom scenarios, simulate interactions, and evaluate the safety and performance of their agents.

Generative Artificial Intelligence (GenAI) constitutes a transformative technological wave that reconfigures industries through its unparalleled capabilities for content creation, reasoning, planning, and multimodal understanding. This revolutionary force offers the most promising path yet toward solving one of engineering's grandest challenges: achieving reliable, fully autonomous driving, particularly the pursuit of Level 5 autonomy. This survey delivers a comprehensive and critical synthesis of the emerging role of GenAI across the autonomous driving stack. We begin by distilling the principles and trade-offs of modern generative modeling, encompassing VAEs, GANs, Diffusion Models, and Large Language Models (LLMs). We then map their frontier applications in image, LiDAR, trajectory, occupancy, video generation as well as LLM-guided reasoning and decision making. We categorize practical applications, such as synthetic data workflows, end-to-end driving strategies, high-fidelity digital twin systems, smart transportation networks, and cross-domain transfer to embodied AI. We identify key obstacles and possibilities such as comprehensive generalization across rare cases, evaluation and safety checks, budget-limited implementation, regulatory compliance, ethical concerns, and environmental effects, while proposing research plans across theoretical assurances, trust metrics, transport integration, and socio-technical influence. By unifying these threads, the survey provides a forward-looking reference for researchers, engineers, and policymakers navigating the convergence of generative AI and advanced autonomous mobility. An actively maintained repository of cited works is available at https://github.com/taco-group/GenAI4AD.

AI coding agents now read repositories, call tools, and execute shell commands with limited human oversight, and a fast-growing body of work studies whether the execution layer around them is actually safe. That literature is scattered. Papers on sandbox isolation, capability and access control, policy enforcement, time-of-check-to-time-of-use (TOCTOU) races, Model Context Protocol (MCP) threats, identity delegation, execution provenance, network egress control, and static analysis of agent-generated code are published independently and rarely cite one another. We systematize 39 papers published between 2023 and 2026 into 17 categories, each verified directly against its source. The same verification protocol also confirms four disclosed, patched CVEs directly affecting production agent harnesses. Reading across categories surfaces five cross-cutting gaps that no single paper addresses. (1) Isolation architectures and capability models are almost never evaluated against one another on a shared benchmark. (2) Policy-enforcement studies report failure rates from 69% to 98% of real denylists, yet no isolation paper re-evaluates its own defense under that adversarial setting. (3) TOCTOU and MCP threats are analyzed as separate literatures despite both being instances of the same state-validation problem. (4) Every enforcement mechanism assumes an honest policy author, leaving policy-authoring error itself unaddressed. (5) Benign but out-of-scope agent actions occurring at rates up to 17.1% under realistic prompting are addressed by no access-control or capability paper in the corpus. Existing broader surveys of agentic AI security discuss sandboxing only as one item among many defenses, leaving execution security without a dedicated systematization. This paper is written to fill that gap. We conclude with a research agenda directed at the five gaps.

This paper introduces a dynamic and actionable framework for securing agentic AI systems in enterprise deployment. We contend that safety and security are not merely fixed attributes of individual models but also emergent properties arising from the dynamic interactions among models, orchestrators, tools, and data within their operating environments. We propose a new way of identification of novel agentic risks through the lens of user safety. Although, for traditional LLMs and agentic models in isolation, safety and security has a clear separation, through the lens of safety in agentic systems, they appear to be connected. Building on this foundation, we define an operational agentic risk taxonomy that unifies traditional safety and security concerns with novel, uniquely agentic risks, including tool misuse, cascading action chains, and unintended control amplification among others. At the core of our approach is a dynamic agentic safety and security framework that operationalizes contextual agentic risk management by using auxiliary AI models and agents, with human oversight, to assist in contextual risk discovery, evaluation, and mitigation. We further address one of the most challenging aspects of safety and security of agentic systems: risk discovery through sandboxed, AI-driven red teaming. We demonstrate the framework effectiveness through a detailed case study of NVIDIA flagship agentic research assistant, AI-Q Research Assistant, showcasing practical, end-to-end safety and security evaluations in complex, enterprise-grade agentic workflows. This risk discovery phase finds novel agentic risks that are then contextually mitigated. We also release the dataset from our case study, containing traces of over 10,000 realistic attack and defense executions of the agentic workflow to help advance research in agentic safety.

Language models can distinguish between testing and deployment phases -- a capability known as evaluation awareness. This has significant safety and policy implications, potentially undermining the reliability of evaluations that are central to AI governance frameworks and voluntary industry commitments. In this paper, we study evaluation awareness in Llama-3.3-70B-Instruct. We show that linear probes can separate real-world evaluation and deployment prompts, suggesting that current models internally represent this distinction. We also find that current safety evaluations are correctly classified by the probes, suggesting that they already appear artificial or inauthentic to models. Our findings underscore the importance of ensuring trustworthy evaluations and understanding deceptive capabilities. More broadly, our work showcases how model internals may be leveraged to support blackbox methods in safety audits, especially for future models more competent at evaluation awareness and deception.

Current AI agents can flexibly invoke tools and execute complex tasks, yet their long-term advancement is hindered by the lack of systematic accumulation and transfer of skills. Without a unified mechanism for skill consolidation, agents frequently ``reinvent the wheel'', rediscovering solutions in isolated contexts without leveraging prior strategies. To overcome this limitation, we introduce SkillNet, an open infrastructure designed to create, evaluate, and organize AI skills at scale. SkillNet structures skills within a unified ontology that supports creating skills from heterogeneous sources, establishing rich relational connections, and performing multi-dimensional evaluation across Safety, Completeness, Executability, Maintainability, and Cost-awareness. Our infrastructure integrates a repository of over 200,000 skills, an interactive platform, and a versatile Python toolkit. Experimental evaluations on ALFWorld, WebShop, and ScienceWorld demonstrate that SkillNet significantly enhances agent performance, improving average rewards by 40% and reducing execution steps by 30% across multiple backbone models. By formalizing skills as evolving, composable assets, SkillNet provides a robust foundation for agents to move from transient experience to durable mastery.

Large language models (LLMs) are increasingly proposed for aviation business operations, from documentation and training generation to customer facing assistants. General purpose benchmarks do not measure whether a model reasons safely and correctly about aviation specific operational knowledge, and the high stakes, regulated nature of the domain makes that gap consequential. We present Pre-Flight, an open source benchmark of 300 multiple choice questions drawn from international standards and airport ground operations material, covering international airport ground operations, ICAO and US FAA regulations, aviation general knowledge and complex operational scenarios. Questions were authored and reviewed by practitioners with experience in air traffic management, ground operations and commercial flying. We evaluate a range of contemporary commercial and open weight models using the Inspect evaluation framework, scoring by accuracy under a standard multiple choice protocol, and we maintain the leaderboard on a rolling basis as new models are released. Against an informal expert reference of around 95%, obtained from a low sample quiz of aviation professionals at a conference, even the strongest model evaluated (released in 2026) reaches 82.7%, having improved only gradually from roughly 75% in early 2025. A substantial and persistent gap below expert level reliability therefore remains. We release the dataset, the evaluation harness and the results, and the benchmark is available within the community evaluations package distributed with inspect_evals. We argue that domain specific evaluation of this kind is a necessary precondition for responsible deployment of generative AI in non safety critical aviation operations.

The advent of large language models (LLMs) has catalyzed a transformative shift in artificial intelligence, paving the way for advanced intelligent agents capable of sophisticated reasoning, robust perception, and versatile action across diverse domains. As these agents increasingly drive AI research and practical applications, their design, evaluation, and continuous improvement present intricate, multifaceted challenges. This survey provides a comprehensive overview, framing intelligent agents within a modular, brain-inspired architecture that integrates principles from cognitive science, neuroscience, and computational research. We structure our exploration into four interconnected parts. First, we delve into the modular foundation of intelligent agents, systematically mapping their cognitive, perceptual, and operational modules onto analogous human brain functionalities, and elucidating core components such as memory, world modeling, reward processing, and emotion-like systems. Second, we discuss self-enhancement and adaptive evolution mechanisms, exploring how agents autonomously refine their capabilities, adapt to dynamic environments, and achieve continual learning through automated optimization paradigms, including emerging AutoML and LLM-driven optimization strategies. Third, we examine collaborative and evolutionary multi-agent systems, investigating the collective intelligence emerging from agent interactions, cooperation, and societal structures, highlighting parallels to human social dynamics. Finally, we address the critical imperative of building safe, secure, and beneficial AI systems, emphasizing intrinsic and extrinsic security threats, ethical alignment, robustness, and practical mitigation strategies necessary for trustworthy real-world deployment.

Safety reasoning is a recent paradigm where LLMs reason over safety policies before generating responses, thereby mitigating limitations in existing safety measures such as over-refusal and jailbreak vulnerabilities. However, implementing this paradigm is challenging due to the resource-intensive process of creating high-quality policy-embedded chain-of-thought (CoT) datasets while ensuring reasoning remains accurate and free from hallucinations or policy conflicts. To tackle this, we propose AIDSAFE: Agentic Iterative Deliberation for Safety Reasoning, a novel data generation recipe that leverages multi-agent deliberation to iteratively expand reasoning on safety policies. A data refiner stage in AIDSAFE ensures high-quality outputs by eliminating repetitive, redundant, and deceptive thoughts. AIDSAFE-generated CoTs provide a strong foundation for supervised fine-tuning (SFT)-based safety training. Additionally, to address the need of preference data in alignment stages, such as DPO training, we introduce a supplemental recipe that uses belief augmentation to create distinct selected and rejected CoT samples. Our evaluations demonstrate that AIDSAFE-generated CoTs achieve superior policy adherence and reasoning quality. Consequently, we show that fine-tuning open-source LLMs on these CoTs can significantly improve safety generalization and jailbreak robustness while maintaining acceptable utility and over-refusal accuracy. AIDSAFE-generated CoT datasets can be found here: https://huggingface.co/datasets/AmazonScience/AIDSAFE

Red-team evaluations of AI models support some claims and not others, and the boundary between the two is calculable rather than merely a matter of judgment. We define the evidential ceiling of an evaluation as the largest factor by which one result can move belief under a fixed testing budget, derive it in closed form for the benchmark null result, and use it to locate that boundary exactly. We find that above a calculable harm rate, a benchmark of modest size certifies a category to a stated evidentiary standard, and a clean sheet is then the stronger of the two possible observations, outweighing a single reproduced failure. Below that rate, no passive benchmark of feasible size provides the specified evidence of safety under the fixed scoring rule and approximately independent trial structure. The crossing between the two regimes has a closed form. The bound is not specific to benchmarks: written in terms of a procedure's hypothesis conditioned elicitation rates, it covers adaptive and automated red teaming as well, and shows that discrimination between the hypotheses rather than attack success is what determines evidential worth. Auditing eight evaluation suites against the boundary, we find that current benchmarks are adequate for high-frequency harm categories and several orders of magnitude short for rare, catastrophic ones. Safety benchmarks are not uninformative. They are informative about a specific and computable set of propositions, and the discipline they need is to state which.

Large Language Models (LLMs) are increasingly used in healthcare, yet ensuring their safety and trustworthiness remains a barrier to deployment. Conversational medical assistants must avoid unsafe compliance without over-refusing benign queries. We present an iterative post-deployment alignment framework that applies Kahneman-Tversky Optimization (KTO) and Direct Preference Optimization (DPO) to refine models against domain-specific safety signals. Using the CARES-18K benchmark for adversarial robustness, we evaluate four LLMs (Llama-3B/8B, Meditron-8B, Mistral-7B) across multiple cycles. Our results show up to 42% improvement in safety-related metrics for harmful query detection, alongside interesting trade-offs against erroneous refusals, thereby exposing architecture-dependent calibration biases. We also perform ablation studies to identify when self-evaluation is reliable and when external or finetuned judges are necessary to maximize performance gains. Our findings underscore the importance of adopting best practices that balance patient safety, user trust, and clinical utility in the design of conversational medical assistants.

This research offers a unique evaluation of how AI systems interpret the digital language of Generation Alpha (Gen Alpha, born 2010-2024). As the first cohort raised alongside AI, Gen Alpha faces new forms of online risk due to immersive digital engagement and a growing mismatch between their evolving communication and existing safety tools. Their distinct language, shaped by gaming, memes, and AI-driven trends, often conceals harmful interactions from both human moderators and automated systems. We assess four leading AI models (GPT-4, Claude, Gemini, and Llama 3) on their ability to detect masked harassment and manipulation within Gen Alpha discourse. Using a dataset of 100 recent expressions from gaming platforms, social media, and video content, the study reveals critical comprehension failures with direct implications for online safety. This work contributes: (1) a first-of-its-kind dataset capturing Gen Alpha expressions; (2) a framework to improve AI moderation systems for youth protection; (3) a multi-perspective evaluation including AI systems, human moderators, and parents, with direct input from Gen Alpha co-researchers; and (4) an analysis of how linguistic divergence increases youth vulnerability. Findings highlight the urgent need to redesign safety systems attuned to youth communication, especially given Gen Alpha reluctance to seek help when adults fail to understand their digital world. This study combines the insight of a Gen Alpha researcher with systematic academic analysis to address critical digital safety challenges.

The proliferation of autonomous AI agents capable of executing real-world actions - filesystem operations, API calls, database modifications, financial transactions - introduces a class of safety risk not addressed by existing content-moderation infrastructure. Current text-safety systems evaluate linguistic content for harm categories such as violence, hate speech, and sexual content; they are architecturally unsuitable for evaluating whether a proposed action falls within an agent's authorized operational scope. We present ILION (Intelligent Logic Identity Operations Network), a deterministic execution gate for agentic AI systems. ILION employs a five-component cascade architecture - Transient Identity Imprint (TII), Semantic Vector Reference Frame (SVRF), Identity Drift Control (IDC), Identity Resonance Score (IRS) and Consensus Veto Layer (CVL) - to classify proposed agent actions as BLOCK or ALLOW without statistical training or API dependencies. The system requires zero labeled data, operates in sub-millisecond latency, and produces fully interpretable verdicts. We evaluate ILION on ILION-Bench v2, a purpose-built benchmark of 380 test scenarios across eight attack categories with 39% hard-difficulty adversarial cases and a held-out development split. ILION achieves F1 = 0.8515, precision = 91.0%, and a false positive rate of 7.9% at a mean latency of 143 microseconds. Comparative evaluation against three baselines - Lakera Guard (F1 = 0.8087), OpenAI Moderation API (F1 = 0.1188), and Llama Guard 3 (F1 = 0.0105) - demonstrates that existing text-safety infrastructure systematically fails on agent execution safety tasks due to a fundamental task mismatch. ILION outperforms the best commercial baseline by 4.3 F1 points while operating 2,000 times faster with a false positive rate four times lower.

As artificial intelligence (AI) models are scaled up, new capabilities can emerge unintentionally and unpredictably, some of which might be dangerous. In response, dangerous capabilities evaluations have emerged as a new risk assessment tool. But what should frontier AI developers do if sufficiently dangerous capabilities are in fact discovered? This paper focuses on one possible response: coordinated pausing. It proposes an evaluation-based coordination scheme that consists of five main steps: (1) Frontier AI models are evaluated for dangerous capabilities. (2) Whenever, and each time, a model fails a set of evaluations, the developer pauses certain research and development activities. (3) Other developers are notified whenever a model with dangerous capabilities has been discovered. They also pause related research and development activities. (4) The discovered capabilities are analyzed and adequate safety precautions are put in place. (5) Developers only resume their paused activities if certain safety thresholds are reached. The paper also discusses four concrete versions of that scheme. In the first version, pausing is completely voluntary and relies on public pressure on developers. In the second version, participating developers collectively agree to pause under certain conditions. In the third version, a single auditor evaluates models of multiple developers who agree to pause if any model fails a set of evaluations. In the fourth version, developers are legally required to run evaluations and pause if dangerous capabilities are discovered. Finally, the paper discusses the desirability and feasibility of our proposed coordination scheme. It concludes that coordinated pausing is a promising mechanism for tackling emerging risks from frontier AI models. However, a number of practical and legal obstacles need to be overcome, especially how to avoid violations of antitrust law.

This report presents a comparative evaluation of DKnownAI Guard in AI agent security scenarios, benchmarked against three competing products: AWS Bedrock Guardrails, Azure Content Safety, and Lakera Guard. Using human annotation as the ground truth, we assess each guardrail's ability to detect two categories of risks: threats to the agent itself (e.g., instruction override, indirect injection, tool abuse) and requests intended to elicit harmful content (e.g., hate speech, pornography, violence). Evaluation results demonstrate that DKnownAI Guard achieves the highest recall rate at 96.5\% and ranks first in true negative rate (TNR) at 90.4\%, delivering the best overall performance among all evaluated guardrails.

To generate evidence regarding the safety and efficacy of artificial intelligence (AI) enabled medical devices, AI models need to be evaluated on a diverse population of patient cases, some of which may not be readily available. We propose an evaluation approach for testing medical imaging AI models that relies on in silico imaging pipelines in which stochastic digital models of human anatomy (in object space) with and without pathology are imaged using a digital replica imaging acquisition system to generate realistic synthetic image datasets. Here, we release M-SYNTH, a dataset of cohorts with four breast fibroglandular density distributions imaged at different exposure levels using Monte Carlo x-ray simulations with the publicly available Virtual Imaging Clinical Trial for Regulatory Evaluation (VICTRE) toolkit. We utilize the synthetic dataset to analyze AI model performance and find that model performance decreases with increasing breast density and increases with higher mass density, as expected. As exposure levels decrease, AI model performance drops with the highest performance achieved at exposure levels lower than the nominal recommended dose for the breast type.

The widespread deployment of general-purpose AI (GPAI) systems introduces significant new risks. Yet the infrastructure, practices, and norms for reporting flaws in GPAI systems remain seriously underdeveloped, lagging far behind more established fields like software security. Based on a collaboration between experts from the fields of software security, machine learning, law, social science, and policy, we identify key gaps in the evaluation and reporting of flaws in GPAI systems. We call for three interventions to advance system safety. First, we propose using standardized AI flaw reports and rules of engagement for researchers in order to ease the process of submitting, reproducing, and triaging flaws in GPAI systems. Second, we propose GPAI system providers adopt broadly-scoped flaw disclosure programs, borrowing from bug bounties, with legal safe harbors to protect researchers. Third, we advocate for the development of improved infrastructure to coordinate distribution of flaw reports across the many stakeholders who may be impacted. These interventions are increasingly urgent, as evidenced by the prevalence of jailbreaks and other flaws that can transfer across different providers' GPAI systems. By promoting robust reporting and coordination in the AI ecosystem, these proposals could significantly improve the safety, security, and accountability of GPAI systems.

Image safety classifiers play an important role in identifying and mitigating the spread of unsafe images online (e.g., images including violence, hateful rhetoric, etc.). At the same time, with the advent of text-to-image models and increasing concerns about the safety of AI models, developers are increasingly relying on image safety classifiers to safeguard their models. Yet, the performance of current image safety classifiers remains unknown for real-world and AI-generated images. To bridge this research gap, in this work, we propose UnsafeBench, a benchmarking framework that evaluates the effectiveness and robustness of image safety classifiers. First, we curate a large dataset of 10K real-world and AI-generated images that are annotated as safe or unsafe based on a set of 11 unsafe categories of images (sexual, violent, hateful, etc.). Then, we evaluate the effectiveness and robustness of five popular image safety classifiers, as well as three classifiers that are powered by general-purpose visual language models. Our assessment indicates that existing image safety classifiers are not comprehensive and effective enough in mitigating the multifaceted problem of unsafe images. Also, we find that classifiers trained only on real-world images tend to have degraded performance when applied to AI-generated images. Motivated by these findings, we design and implement a comprehensive image moderation tool called PerspectiveVision, which effectively identifies 11 categories of real-world and AI-generated unsafe images. The best PerspectiveVision model achieves an overall F1-Score of 0.810 on six evaluation datasets, which is comparable with closed-source and expensive state-of-the-art models like GPT-4V. UnsafeBench and PerspectiveVision can aid the research community in better understanding the landscape of image safety classification in the era of generative AI.

As humans delegate more tasks and decisions to artificial intelligence (AI), we risk losing control of our individual and collective futures. Relatively simple algorithmic systems already steer human decision-making, such as social media feed algorithms that lead people to unintentionally and absent-mindedly scroll through engagement-optimized content. In this paper, we develop the idea of human agency by integrating philosophical and scientific theories of agency with AI-assisted evaluation methods: using large language models (LLMs) to simulate and validate user queries and to evaluate AI responses. We develop HumanAgencyBench (HAB), a scalable and adaptive benchmark with six dimensions of human agency based on typical AI use cases. HAB measures the tendency of an AI assistant or agent to Ask Clarifying Questions, Avoid Value Manipulation, Correct Misinformation, Defer Important Decisions, Encourage Learning, and Maintain Social Boundaries. We find low-to-moderate agency support in contemporary LLM-based assistants and substantial variation across system developers and dimensions. For example, while Anthropic LLMs most support human agency overall, they are the least supportive LLMs in terms of Avoid Value Manipulation. Agency support does not appear to consistently result from increasing LLM capabilities or instruction-following behavior (e.g., RLHF), and we encourage a shift towards more robust safety and alignment targets.

Global safety models exhibit strong performance across widely used benchmarks, yet their training data rarely captures the cultural and linguistic nuances of Taiwanese Mandarin. This limitation results in systematic blind spots when interpreting region-specific risks such as localized financial scams, culturally embedded hate speech, and misinformation patterns. To address these gaps, we introduce TS-Bench (Taiwan Safety Benchmark), a standardized evaluation suite for assessing safety performance in Taiwanese Mandarin. TS-Bench contains 400 human-curated prompts spanning critical domains including financial fraud, medical misinformation, social discrimination, and political manipulation. In parallel, we present Breeze Guard, an 8B safety model derived from Breeze 2, our previously released general-purpose Taiwanese Mandarin LLM with strong cultural grounding from its original pre-training corpus. Breeze Guard is obtained through supervised fine-tuning on a large-scale, human-verified synthesized dataset targeting Taiwan-specific harms. Our central hypothesis is that effective safety detection requires the cultural grounding already present in the base model; safety fine-tuning alone is insufficient to introduce new socio linguistic knowledge from scratch. Empirically, Breeze Guard significantly outperforms the leading 8B general-purpose safety model, Granite Guardian 3.3, on TS-Bench (+0.17 overall F1), with particularly large gains in high-context categories such as scam (+0.66 F1) and financial malpractice (+0.43 F1). While the model shows slightly lower performance on English-centric benchmarks (ToxicChat, AegisSafetyTest), this tradeoff is expected for a regionally specialized safety model optimized for Taiwanese Mandarin. Together, Breeze Guard and TS-Bench establish a new foundation for trustworthy AI deployment in Taiwan.

Transportation safety analysis requires integrating crash records, roadway attributes, and geospatial data through GIS-based workflows, but access remains uneven across agencies and community stakeholders. Technical prerequisites create a gap between analytical tools central to safety planning and the practitioners able to use them. Local agencies, school committees, and residents may have safety concerns but limited capacity to retrieve, filter, map, and analyze relevant data. Generative AI offers a way to narrow this divide, but its public-sector use raises questions about reliability, reproducibility, and governance. This paper presents a schema-grounded natural language interface for transportation safety analysis, using a large language model (LLM) to interpret user intent while preserving deterministic, reviewable execution against an authoritative database. User queries are translated into structured semantic frames, validated by a rule-based layer, compiled into a typed directed acyclic graph of spatial operations, and executed against a PostGIS database. This bounded design separates language interpretation from deterministic execution, keeping results reproducible and schema-grounded while removing access barriers. The framework is evaluated using a statewide Massachusetts transportation safety database integrating crash records, roadway attributes, and geospatial layers including schools, bus stops, crosswalks, and municipal boundaries. All queries executed successfully; the validation layer corrects errors in 29% of evaluation queries, reflecting the gap between flexible natural language and strict schema-grounded requirements. The results suggest that combining natural language accessibility with deterministic execution is a practical direction for broadening access to transportation safety data, with implications for trustworthy AI in public-sector planning.

As reasoning capacity and deployment scope grow in tandem, large language models (LLMs) gain the capacity to engage in behaviors that serve their own objectives, a class of risks we term Emergent Strategic Reasoning Risks (ESRRs). These include, but are not limited to, deception (intentionally misleading users or evaluators), evaluation gaming (strategically manipulating performance during safety testing), and reward hacking (exploiting misspecified objectives). Systematically understanding and benchmarking these risks remains an open challenge. To address this gap, we introduce ESRRSim, a taxonomy-driven agentic framework for automated behavioral risk evaluation. We construct an extensible risk taxonomy of 7 categories, which is decomposed into 20 subcategories. ESRRSim generates evaluation scenarios designed to elicit faithful reasoning, paired with dual rubrics assessing both model responses and reasoning traces, in a judge-agnostic and scalable architecture. Evaluation across 11 reasoning LLMs reveals substantial variation in risk profiles (detection rates ranging 14.45%-72.72%), with dramatic generational improvements suggesting models may increasingly recognize and adapt to evaluation contexts.

Large language models (LLMs) are increasingly explored as scalable tools for mental health counseling, yet evaluating their safety remains challenging due to the interactional and context-dependent nature of clinical harm. Existing evaluation frameworks predominantly assess isolated responses using coarse-grained taxonomies or static datasets, limiting their ability to diagnose how harms emerge and accumulate over multi-turn counseling interactions. In this work, we introduce R-MHSafe, a role-aware mental health safety taxonomy that characterizes clinically significant harm in terms of the interactional roles an AI counselor adopts, including perpetrator, instigator, facilitator, or enabler, combined with clinically grounded harm categories. Then, we propose MHSafeEval, a closed-loop, agent-based evaluation framework that formulates safety assessment as trajectory-level discovery of harm through adversarial multi-turn interactions, guided by role-aware modeling. Using R-MHSafe and MHSafeEval, we conduct a large-scale evaluation across state-of-the-art LLMs. Our results reveal substantial role-dependent and cumulative safety failures that are systematically missed by existing static benchmarks, and show that our framework significantly improves failure-mode coverage and diagnostic granularity.

To address this gap, we introduce Libra-Leaderboard, a comprehensive framework designed to rank LLMs through a balanced evaluation of performance and safety. Combining a dynamic leaderboard with an interactive LLM arena, Libra-Leaderboard encourages the joint optimization of capability and safety. Unlike traditional approaches that average performance and safety metrics, Libra-Leaderboard uses a distance-to-optimal-score method to calculate the overall rankings. This approach incentivizes models to achieve a balance rather than excelling in one dimension at the expense of some other ones. In the first release, Libra-Leaderboard evaluates 26 mainstream LLMs from 14 leading organizations, identifying critical safety challenges even in state-of-the-art models.

Evaluating generative AI models is increasingly resource-intensive due to slow inference, expensive raters, and a rapidly growing landscape of models and benchmarks. We propose ProEval, a proactive evaluation framework that leverages transfer learning to efficiently estimate performance and identify failure cases. ProEval employs pre-trained Gaussian Processes (GPs) as surrogates for the performance score function, mapping model inputs to metrics such as the severity of errors or safety violations. By framing performance estimation as Bayesian quadrature (BQ) and failure discovery as superlevel set sampling, we develop uncertainty-aware decision strategies that actively select or synthesize highly informative inputs for testing. Theoretically, we prove that our pre-trained GP-based BQ estimator is unbiased and bounded. Empirically, extensive experiments on reasoning, safety alignment, and classification benchmarks demonstrate that ProEval is significantly more efficient than competitive baselines. It requires 8-65x fewer samples to achieve estimates within 1% of the ground truth, while simultaneously revealing more diverse failure cases under a stricter evaluation budget.

AI agents -- systems that combine foundation models with reasoning, planning, memory, and tool use -- are rapidly becoming a practical interface between natural-language intent and real-world computation. This survey synthesizes the emerging landscape of AI agent architectures across: (i) deliberation and reasoning (e.g., chain-of-thought-style decomposition, self-reflection and verification, and constraint-aware decision making), (ii) planning and control (from reactive policies to hierarchical and multi-step planners), and (iii) tool calling and environment interaction (retrieval, code execution, APIs, and multimodal perception). We organize prior work into a unified taxonomy spanning agent components (policy/LLM core, memory, world models, planners, tool routers, and critics), orchestration patterns (single-agent vs.\ multi-agent; centralized vs.\ decentralized coordination), and deployment settings (offline analysis vs.\ online interactive assistance; safety-critical vs.\ open-ended tasks). We discuss key design trade-offs -- latency vs.\ accuracy, autonomy vs.\ controllability, and capability vs.\ reliability -- and highlight how evaluation is complicated by non-determinism, long-horizon credit assignment, tool and environment variability, and hidden costs such as retries and context growth. Finally, we summarize measurement and benchmarking practices (task suites, human preference and utility metrics, success under constraints, robustness and security) and identify open challenges including verification and guardrails for tool actions, scalable memory and context management, interpretability of agent decisions, and reproducible evaluation under realistic workloads.

Vision-Language-Action (VLA) models are emerging as a unified substrate for embodied intelligence. This shift raises a new class of safety challenges, stemming from the embodied nature of VLA systems, including irreversible physical consequences, a multimodal attack surface across vision, language, and state, real-time latency constraints on defense, error propagation over long-horizon trajectories, and vulnerabilities in the data supply chain. Yet the literature remains fragmented across robotic learning, adversarial machine learning, AI alignment, and autonomous systems safety. This survey provides a unified and up-to-date overview of safety in Vision-Language-Action models. We organize the field along two parallel timing axes, attack timing (training-time vs. inference-time and defense timing (training-time vs. inference-time, linking each class of threat to the stage at which it can be mitigated. We first define the scope of VLA safety, distinguishing it from text-only LLM safety and classical robotic safety, and review the foundations of VLA models, including architectures, training paradigms, and inference mechanisms. We then examine the literature through four lenses: Attacks, Defenses, Evaluation, and Deployment. We survey training-time threats such as data poisoning and backdoors, as well as inference-time attacks including adversarial patches, cross-modal perturbations, semantic jailbreaks, and freezing attacks. We review training-time and runtime defenses, analyze existing benchmarks and metrics, and discuss safety challenges across six deployment domains. Finally, we highlight key open problems, including certified robustness for embodied trajectories, physically realizable defenses, safety-aware training, unified runtime safety architectures, and standardized evaluation.

For safety, medical AI systems undergo thorough evaluations before deployment, validating their predictions against a ground truth which is assumed to be fixed and certain. However, this ground truth is often curated in the form of differential diagnoses. While a single differential diagnosis reflects the uncertainty in one expert assessment, multiple experts introduce another layer of uncertainty through disagreement. Both forms of uncertainty are ignored in standard evaluation which aggregates these differential diagnoses to a single label. In this paper, we show that ignoring uncertainty leads to overly optimistic estimates of model performance, therefore underestimating risk associated with particular diagnostic decisions. To this end, we propose a statistical aggregation approach, where we infer a distribution on probabilities of underlying medical condition candidates themselves, based on observed annotations. This formulation naturally accounts for the potential disagreements between different experts, as well as uncertainty stemming from individual differential diagnoses, capturing the entire ground truth uncertainty. Our approach boils down to generating multiple samples of medical condition probabilities, then evaluating and averaging performance metrics based on these sampled probabilities. In skin condition classification, we find that a large portion of the dataset exhibits significant ground truth uncertainty and standard evaluation severely over-estimates performance without providing uncertainty estimates. In contrast, our framework provides uncertainty estimates on common metrics of interest such as top-k accuracy and average overlap, showing that performance can change multiple percentage points. We conclude that, while assuming a crisp ground truth can be acceptable for many AI applications, a more nuanced evaluation protocol should be utilized in medical diagnosis.

Safety alignment in large language models relies predominantly on English-language training data. When harmful intent is expressed in low-resource languages, refusal mechanisms that hold in English frequently fail to activate. We introduce LSR (Linguistic Safety Robustness), the first systematic benchmark for measuring cross-lingual refusal degradation in West African languages: Yoruba, Hausa, Igbo, and Igala. LSR uses a dual-probe evaluation protocol - submitting matched English and target-language probes to the same model - and introduces Refusal Centroid Drift (RCD), a metric that quantifies how much of a model's English refusal behavior is lost when harmful intent is encoded in a target language. We evaluate Gemini 2.5 Flash across 14 culturally grounded attack probes in four harm categories. English refusal rates hold at approximately 90 percent. Across West African languages, refusal rates fall to 35-55 percent, with Igala showing the most severe degradation (RCD = 0.55). LSR is implemented in the Inspect AI evaluation framework and is available as a PR-ready contribution to the UK AISI's inspect_evals repository. A live reference implementation and the benchmark dataset are publicly available.

Scams exploiting real-time social engineering -- such as phishing, impersonation, and phone fraud -- remain a persistent and evolving threat across digital platforms. Existing defenses are largely reactive, offering limited protection during active interactions. We propose a privacy-preserving, AI-in-the-loop framework that proactively detects and disrupts scam conversations in real time. The system combines instruction-tuned artificial intelligence with a safety-aware utility function that balances engagement with harm minimization, and employs federated learning to enable continual model updates without raw data sharing. Experimental evaluations show that the system produces fluent and engaging responses (perplexity as low as 22.3, engagement approx0.80), while human studies confirm significant gains in realism, safety, and effectiveness over strong baselines. In federated settings, models trained with FedAvg sustain up to 30 rounds while preserving high engagement (approx0.80), strong relevance (approx0.74), and low PII leakage (leq0.0085). Even with differential privacy, novelty and safety remain stable, indicating that robust privacy can be achieved without sacrificing performance. The evaluation of guard models (LlamaGuard, LlamaGuard2/3, MD-Judge) shows a straightforward pattern: stricter moderation settings reduce the chance of exposing personal information, but they also limit how much the model engages in conversation. In contrast, more relaxed settings allow longer and richer interactions, which improve scam detection, but at the cost of higher privacy risk. To our knowledge, this is the first framework to unify real-time scam-baiting, federated privacy preservation, and calibrated safety moderation into a proactive defense paradigm.

Large Language Models (LLMs) are increasingly utilized for mental health support; however, current safety benchmarks often fail to detect the complex, longitudinal risks inherent in therapeutic dialogue. We introduce an evaluation framework that pairs AI psychotherapists with simulated patient agents equipped with dynamic cognitive-affective models and assesses therapy session simulations against a comprehensive quality of care and risk ontology. We apply this framework to a high-impact test case, Alcohol Use Disorder, evaluating six AI agents (including ChatGPT, Gemini, and Character.AI) against a clinically-validated cohort of 15 patient personas representing diverse clinical phenotypes. Our large-scale simulation (N=369 sessions) reveals critical safety gaps in the use of AI for mental health support. We identify specific iatrogenic risks, including the validation of patient delusions ("AI Psychosis") and failure to de-escalate suicide risk. Finally, we validate an interactive data visualization dashboard with diverse stakeholders, including AI engineers and red teamers, mental health professionals, and policy experts (N=9), demonstrating that this framework effectively enables stakeholders to audit the "black box" of AI psychotherapy. These findings underscore the critical safety risks of AI-provided mental health support and the necessity of simulation-based clinical red teaming before deployment.

AI agents are increasingly deployed to execute important tasks. While rising accuracy scores on standard benchmarks suggest rapid progress, many agents still continue to fail in practice. This discrepancy highlights a fundamental limitation of current evaluations: compressing agent behavior into a single success metric obscures critical operational flaws. Notably, it ignores whether agents behave consistently across runs, withstand perturbations, fail predictably, or have bounded error severity. Grounded in safety-critical engineering, we provide a holistic performance profile by proposing twelve concrete metrics that decompose agent reliability along four key dimensions: consistency, robustness, predictability, and safety. Evaluating 14 agentic models across two complementary benchmarks, we find that recent capability gains have only yielded small improvements in reliability. By exposing these persistent limitations, our metrics complement traditional evaluations while offering tools for reasoning about how agents perform, degrade, and fail.

Adversarial testing of large language models (LLMs) is crucial for their safe and responsible deployment. We introduce a novel approach for automated generation of adversarial evaluation datasets to test the safety of LLM generations on new downstream applications. We call it AI-assisted Red-Teaming (AART) - an automated alternative to current manual red-teaming efforts. AART offers a data generation and augmentation pipeline of reusable and customizable recipes that reduce human effort significantly and enable integration of adversarial testing earlier in new product development. AART generates evaluation datasets with high diversity of content characteristics critical for effective adversarial testing (e.g. sensitive and harmful concepts, specific to a wide range of cultural and geographic regions and application scenarios). The data generation is steered by AI-assisted recipes to define, scope and prioritize diversity within the application context. This feeds into a structured LLM-generation process that scales up evaluation priorities. Compared to some state-of-the-art tools, AART shows promising results in terms of concept coverage and data quality.

People increasingly seek personal advice from large language models (LLMs), yet whether humans follow their advice, and its consequences for their well-being, remains unknown. In a longitudinal randomised controlled trial with a representative UK sample (N = 2,302), 75% of participants who had a 20-minute discussion with GPT-4o about health, careers or relationships subsequently reported following its advice. Based on autograder evaluations of chat transcripts, LLM advice rarely violated safety best practice. When queried 2-3 weeks later, participants who had interacted with personalised AI (with access to detailed user information) followed its advice more often in the real world and reported higher well-being than those advised by non-personalised AI. However, while receiving personal advice from AI temporarily reduced well-being, no differential long-term effects compared to a control emerged. Our results suggest that humans readily follow LLM advice about personal issues but doing so shows no additional well-being benefit over casual conversations.

Large language models (LLMs) show promise in clinical decision support yet risk acquiescing to patient pressure for inappropriate care. We introduce SycoEval-EM, a multi-agent simulation framework evaluating LLM robustness through adversarial patient persuasion in emergency medicine. Across 20 LLMs and 1,875 encounters spanning three Choosing Wisely scenarios, acquiescence rates ranged from 0-100\%. Models showed higher vulnerability to imaging requests (38.8\%) than opioid prescriptions (25.0\%), with model capability poorly predicting robustness. All persuasion tactics proved equally effective (30.0-36.0\%), indicating general susceptibility rather than tactic-specific weakness. Our findings demonstrate that static benchmarks inadequately predict safety under social pressure, necessitating multi-turn adversarial testing for clinical AI certification.

The rapid advancement of large models, driven by their exceptional abilities in learning and generalization through large-scale pre-training, has reshaped the landscape of Artificial Intelligence (AI). These models are now foundational to a wide range of applications, including conversational AI, recommendation systems, autonomous driving, content generation, medical diagnostics, and scientific discovery. However, their widespread deployment also exposes them to significant safety risks, raising concerns about robustness, reliability, and ethical implications. This survey provides a systematic review of current safety research on large models, covering Vision Foundation Models (VFMs), Large Language Models (LLMs), Vision-Language Pre-training (VLP) models, Vision-Language Models (VLMs), Diffusion Models (DMs), and large-model-based Agents. Our contributions are summarized as follows: (1) We present a comprehensive taxonomy of safety threats to these models, including adversarial attacks, data poisoning, backdoor attacks, jailbreak and prompt injection attacks, energy-latency attacks, data and model extraction attacks, and emerging agent-specific threats. (2) We review defense strategies proposed for each type of attacks if available and summarize the commonly used datasets and benchmarks for safety research. (3) Building on this, we identify and discuss the open challenges in large model safety, emphasizing the need for comprehensive safety evaluations, scalable and effective defense mechanisms, and sustainable data practices. More importantly, we highlight the necessity of collective efforts from the research community and international collaboration. Our work can serve as a useful reference for researchers and practitioners, fostering the ongoing development of comprehensive defense systems and platforms to safeguard AI models.

Despite the growing use of large language models (LLMs) in clinical dialogue systems, existing evaluations focus on task completion or fluency, offering little insight into the behavioral and risk management requirements essential for safety-critical systems. This paper presents MATRIX (Multi-Agent simulaTion fRamework for safe Interactions and conteXtual clinical conversational evaluation), a structured, extensible framework for safety-oriented evaluation of clinical dialogue agents. MATRIX integrates three components: (1) a safety-aligned taxonomy of clinical scenarios, expected system behaviors and failure modes derived through structured safety engineering methods; (2) BehvJudge, an LLM-based evaluator for detecting safety-relevant dialogue failures, validated against expert clinician annotations; and (3) PatBot, a simulated patient agent capable of producing diverse, scenario-conditioned responses, evaluated for realism and behavioral fidelity with human factors expertise, and a patient-preference study. Across three experiments, we show that MATRIX enables systematic, scalable safety evaluation. BehvJudge with Gemini 2.5-Pro achieves expert-level hazard detection (F1 0.96, sensitivity 0.999), outperforming clinicians in a blinded assessment of 240 dialogues. We also conducted one of the first realism analyses of LLM-based patient simulation, showing that PatBot reliably simulates realistic patient behavior in quantitative and qualitative evaluations. Using MATRIX, we demonstrate its effectiveness in benchmarking five LLM agents across 2,100 simulated dialogues spanning 14 hazard scenarios and 10 clinical domains. MATRIX is the first framework to unify structured safety engineering with scalable, validated conversational AI evaluation, enabling regulator-aligned safety auditing. We release all evaluation tools, prompts, structured scenarios, and datasets.

The swift progress and widespread acceptance of artificial intelligence (AI) systems highlight a pressing requirement to comprehend both the capabilities and potential risks associated with AI. Given the linguistic complexity, cultural richness, and underrepresented status of Arabic in AI research, there is a pressing need to focus on Large Language Models (LLMs) performance and safety for Arabic-related tasks. Despite some progress in their development, there is a lack of comprehensive trustworthiness evaluation benchmarks, which presents a major challenge in accurately assessing and improving the safety of LLMs when prompted in Arabic. In this paper, we introduce AraTrust, the first comprehensive trustworthiness benchmark for LLMs in Arabic. AraTrust comprises 522 human-written multiple-choice questions addressing diverse dimensions related to truthfulness, ethics, safety, physical health, mental health, unfairness, illegal activities, privacy, and offensive language. We evaluated a set of LLMs against our benchmark to assess their trustworthiness. GPT-4 was the most trustworthy LLM, while open-source models, particularly AceGPT 7B and Jais 13B, struggled to achieve a score of 60% in our benchmark.

AI agents are increasingly granted economic agency (executing trades, managing budgets, negotiating contracts, and spawning sub-agents), yet current frameworks gate this agency on capability benchmarks that are empirically uncorrelated with operational robustness. We introduce the Comprehension-Gated Agent Economy (CGAE), a formal architecture in which an agent's economic permissions are upper-bounded by a verified comprehension function derived from adversarial robustness audits. The gating mechanism operates over three orthogonal robustness dimensions: constraint compliance (measured by CDCT), epistemic integrity (measured by DDFT), and behavioral alignment (measured by AGT), with intrinsic hallucination rates serving as a cross-cutting diagnostic. We define a weakest-link gate function that maps robustness vectors to discrete economic tiers, and prove three properties of the resulting system: (1) bounded economic exposure, ensuring maximum financial liability is a function of verified robustness; (2) incentive-compatible robustness investment, showing rational agents maximize profit by improving robustness rather than scaling capability alone; and (3) monotonic safety scaling, demonstrating that aggregate system safety does not decrease as the economy grows. The architecture includes temporal decay and stochastic re-auditing mechanisms that prevent post-certification drift. CGAE provides the first formal bridge between empirical AI robustness evaluation and economic governance, transforming safety from a regulatory burden into a competitive advantage.

The field of explainable artificial intelligence (XAI) aims to uncover the inner workings of complex deep neural models. While being crucial for safety-critical domains, XAI inherently lacks ground-truth explanations, making its automatic evaluation an unsolved problem. We address this challenge by proposing a novel synthetic vision dataset, named FunnyBirds, and accompanying automatic evaluation protocols. Our dataset allows performing semantically meaningful image interventions, e.g., removing individual object parts, which has three important implications. First, it enables analyzing explanations on a part level, which is closer to human comprehension than existing methods that evaluate on a pixel level. Second, by comparing the model output for inputs with removed parts, we can estimate ground-truth part importances that should be reflected in the explanations. Third, by mapping individual explanations into a common space of part importances, we can analyze a variety of different explanation types in a single common framework. Using our tools, we report results for 24 different combinations of neural models and XAI methods, demonstrating the strengths and weaknesses of the assessed methods in a fully automatic and systematic manner.

Recent advances in large language models have sparked growing interest in AI agents capable of solving complex, real-world tasks. However, most existing agent systems rely on manually crafted configurations that remain static after deployment, limiting their ability to adapt to dynamic and evolving environments. To this end, recent research has explored agent evolution techniques that aim to automatically enhance agent systems based on interaction data and environmental feedback. This emerging direction lays the foundation for self-evolving AI agents, which bridge the static capabilities of foundation models with the continuous adaptability required by lifelong agentic systems. In this survey, we provide a comprehensive review of existing techniques for self-evolving agentic systems. Specifically, we first introduce a unified conceptual framework that abstracts the feedback loop underlying the design of self-evolving agentic systems. The framework highlights four key components: System Inputs, Agent System, Environment, and Optimisers, serving as a foundation for understanding and comparing different strategies. Based on this framework, we systematically review a wide range of self-evolving techniques that target different components of the agent system. We also investigate domain-specific evolution strategies developed for specialised fields such as biomedicine, programming, and finance, where optimisation objectives are tightly coupled with domain constraints. In addition, we provide a dedicated discussion on the evaluation, safety, and ethical considerations for self-evolving agentic systems, which are critical to ensuring their effectiveness and reliability. This survey aims to provide researchers and practitioners with a systematic understanding of self-evolving AI agents, laying the foundation for the development of more adaptive, autonomous, and lifelong agentic systems.

We present Test-Driven AI Agent Definition (TDAD), a methodology that treats agent prompts as compiled artifacts: engineers provide behavioral specifications, a coding agent converts them into executable tests, and a second coding agent iteratively refines the prompt until tests pass. Deploying tool-using LLM agents in production requires measurable behavioral compliance that current development practices cannot provide. Small prompt changes cause silent regressions, tool misuse goes undetected, and policy violations emerge only after deployment. To mitigate specification gaming, TDAD introduces three mechanisms: (1) visible/hidden test splits that withhold evaluation tests during compilation, (2) semantic mutation testing via a post-compilation agent that generates plausible faulty prompt variants, with the harness measuring whether the test suite detects them, and (3) spec evolution scenarios that quantify regression safety when requirements change. We evaluate TDAD on SpecSuite-Core, a benchmark of four deeply-specified agents spanning policy compliance, grounded analytics, runbook adherence, and deterministic enforcement. Across 24 independent trials, TDAD achieves 92% v1 compilation success with 97% mean hidden pass rate; evolved specifications compile at 58%, with most failed runs passing all visible tests except 1-2, and show 86-100% mutation scores, 78% v2 hidden pass rate, and 97% regression safety scores. The implementation is available as an open benchmark at https://github.com/f-labs-io/tdad-paper-code.

Physical AI systems increasingly map multimodal observations, language instructions, and learned world representations into physically consequential actions. Robotics foundation models, vision-language-action models, and world-model-based autonomous systems can condition decisions that move vehicles, robots, drones, and industrial machines. This transition exposes a safety problem that is not fully captured by conventional AI content moderation or by classical robot safety alone: a black-box model may issue a physically consequential action while appearing confident, plausible, and semantically aligned. The resulting failure can be silent, arising from sensor drift, occlusion, state-estimation error, distribution shift, hallucinated affordances, or invalid physical assumptions before downstream hardware controllers detect a violation. Across embodied foundation models, world models, robotics simulation, embodied safety benchmarks, safe control, runtime assurance, uncertainty estimation, verification, and guardrail evaluation, model capability and safety mechanisms have advanced along largely separate technical tracks. A recurring gap synthesized here is that no single stream surveyed in this review supplies a complete runtime authorization boundary between black-box Physical AI models and physical execution. The resulting analysis develops a bounded problem formulation, a definition of silent physical-action failure, a taxonomy of runtime guardrail functions, and evaluation requirements for comparing guardrails as Physical AI assurance mechanisms.

AI safety has a missing layer. Interpretability, formal methods, security engineering, evaluation methodology, and reinforcement-learning safety each produce substantial work, but the resulting artifacts do not compose into deployable oversight: every team fielding an agentic system builds its own audit schema, policy dialect, monitoring stack, and escalation path, mostly reinventions of patterns understood elsewhere. We diagnose this as a coordination gap, not a research gap, and propose a two-axis taxonomy: five technical layers (Legibility, Specification, Mediation, Evaluation, Escalation) crossed with six concerns spanning alignment, robustness, adversarial defense, security, governance, and accountability, populating the resulting 5x6 matrix with existing work. Layer 2 (Specification), where humans translate intent into machine-checkable artifacts, is the connective tissue every layer depends on, yet it lacks four marks of a mature engineering discipline: shared vocabulary, design principles, composability standards, and governance practices. We propose six design principles for Layer 2, from elicitability and composability to adversary-awareness, traceability, and governability, made concrete through worked examples and a reference architecture turning specifications into runtime enforcement, evaluation, and escalation. Existing systems such as Cedar, Constitutional AI, and Open Policy Agent each address a fragment of Layer 2 well and the matrix poorly; treating them as fragments of one shared layer makes composition tractable. As evidence, we introduce CARMA, a Layer 2 prototype for autonomous ETL agents in which one specification drives enforcement, evaluation, and escalation, with every decision traceable to a versioned specification, naming what AI oversight is missing and giving independent teams principles to build the missing pieces so they compose.

The rapid integration of agentic AI into high-stakes real-world applications requires robust oversight mechanisms. The emerging field of AI Control (AIC) aims to provide such an oversight mechanism, but practical adoption depends heavily on implementation overhead. To study this problem better, we introduce the notion of Control tax -- the operational and financial cost of integrating control measures into AI pipelines. Our work makes three key contributions to the field of AIC: (1) we introduce a theoretical framework that quantifies the Control Tax and maps classifier performance to safety assurances; (2) we conduct comprehensive evaluations of state-of-the-art language models in adversarial settings, where attacker models insert subtle backdoors into code while monitoring models attempt to detect these vulnerabilities; and (3) we provide empirical financial cost estimates for control protocols and develop optimized monitoring strategies that balance safety and cost-effectiveness while accounting for practical constraints like auditing budgets. Our framework enables practitioners to make informed decisions by systematically connecting safety guarantees with their costs, advancing AIC through principled economic feasibility assessment across different deployment contexts.

Large language models (LLMs) are increasingly applied in mental health support systems, where reliable recognition of high-risk states such as suicidal ideation and self-harm is safety-critical. However, existing evaluations primarily rely on aggregate performance metrics, which often obscure risk-specific failure modes and provide limited insight into model behavior in realistic, multi-turn interactions. We present MHDash, an open-source platform designed to support the development, evaluation, and auditing of AI systems for mental health applications. MHDash integrates data collection, structured annotation, multi-turn dialogue generation, and baseline evaluation into a unified pipeline. The platform supports annotations across multiple dimensions, including Concern Type, Risk Level, and Dialogue Intent, enabling fine-grained and risk-aware analysis. Our results reveal several key findings: (i) simple baselines and advanced LLM APIs exhibit comparable overall accuracy yet diverge significantly on high-risk cases; (ii) some LLMs maintain consistent ordinal severity ranking while failing absolute risk classification, whereas others achieve reasonable aggregate scores but suffer from high false negative rates on severe categories; and (iii) performance gaps are amplified in multi-turn dialogues, where risk signals emerge gradually. These observations demonstrate that conventional benchmarks are insufficient for safety-critical mental health settings. By releasing MHDash as an open platform, we aim to promote reproducible research, transparent evaluation, and safety-aligned development of AI systems for mental health support.

Large Language Models (LLMs) often exhibit significant behavioral shifts when they perceive a change from a real-world deployment context to a controlled evaluation setting, a phenomenon known as "evaluation awareness." This discrepancy poses a critical challenge for AI alignment, as benchmark performance may not accurately reflect a model's true safety and honesty. In this work, we systematically quantify these behavioral changes by manipulating the perceived context of prompts. We introduce a methodology that uses a linear probe to score prompts on a continuous scale from "test-like" to "deploy-like" and leverage an LLM rewriting strategy to shift these prompts towards a more natural, deployment-style context while preserving the original task. Using this method, we achieved a 30% increase in the average probe score across a strategic role-playing dataset after rewriting. Evaluating a suite of state-of-the-art models on these original and rewritten prompts, we find that rewritten "deploy-like" prompts induce a significant and consistent shift in behavior. Across all models, we observed an average increase in honest responses of 5.26% and a corresponding average decrease in deceptive responses of 12.40%. Furthermore, refusal rates increased by an average of 6.38%, indicating heightened safety compliance. Our findings demonstrate that evaluation awareness is a quantifiable and manipulable factor that directly influences LLM behavior, revealing that models are more prone to unsafe or deceptive outputs in perceived test environments. This underscores the urgent need for more realistic evaluation frameworks to accurately gauge true model alignment before deployment.

We introduce Phoenix-VL 1.5 Medium, a 123B-parameter natively multimodal and multilingual foundation model, adapted to regional languages and the Singapore context. Developed as a sovereign AI asset, it demonstrates that deep domain adaptation can be achieved with minimal degradation to broad-spectrum intelligence and alignment. Continued pretraining was performed on Mistral Medium 3.1 using a localized 1-trillion tokens multimodal corpus, followed by a 250-billion tokens long-context extension phase. Subsequent post-training incorporated a novel human-annotated Singapore multimodal dataset and curated textual corpus on Singapore culture, knowledge, and legislation, totaling 22-billion tokens. An additional 5 billion tokens of model alignment was performed through Online Direct Preference Optimization. Phoenix-VL 1.5 Medium achieves state-of-the-art performance for its size on Singapore multimodal, legal, and government policy benchmarks while remaining globally competitive on general multimodal intelligence, multilingual, and STEM benchmarks. We also introduce a novel evaluation suite encompassing localized knowledge benchmarks and an institutionally aligned model behavior and safety framework. We report the data curation principles, training methodology, and highlight benchmark and inference performance.

Recent large language models (LLMs) have demonstrated strong capabilities in understanding and generating code, from competitive programming to repository-level software engineering. In emerging agentic systems, code is no longer only a target output. It increasingly serves as an operational substrate for agent reasoning, acting, environment modeling, and execution-based verification. We frame this shift through the lens of agent harnesses and introduce code as agent harness: a unified view that centers code as the basis for agent infrastructure. To systematically study this perspective, we organize the survey around three connected layers. First, we study the harness interface, where code connects agents to reasoning, action, and environment modeling. Second, we examine harness mechanisms: planning, memory, and tool use for long-horizon execution, together with feedback-driven control and optimization that make harness reliable and adaptive. Third, we discuss scaling the harness from single-agent systems to multi-agent settings, where shared code artifacts support multi-agent coordination, review, and verification. Across these layers, we summarize representative methods and practical applications of code as agent harness, spanning coding assistants, GUI/OS automation, embodied agents, scientific discovery, personalization and recommendation, DevOps, and enterprise workflows. We further outline open challenges for harness engineering, including evaluation beyond final task success, verification under incomplete feedback, regression-free harness improvement, consistent shared state across multiple agents, human oversight for safety-critical actions, and extensions to multimodal environments. By centering code as the harness of agentic AI, this survey provides a unified roadmap toward executable, verifiable, and stateful AI agent systems.

Common methods for aligning already-capable models with desired behavior rely on the ability of humans to provide supervision. However, future superhuman models will surpass the capability of humans. Therefore, humans will only be able to weakly supervise superhuman models. This expected deficiency of human evaluation would weaken the safety of future AI systems. Scalable oversight and weak-to-strong generalization are two complementary approaches to tackle this issue. In this paper, we attempt to combine the strengths of these two approaches to further improve alignment. Specifically, we investigate ways of improving human supervision with a strong pretrained model and then supervise the strong model with enhanced weak human supervision. To make iterative empirical progress, we consider an analogy: can we use a strong model to improve weak model supervision and then use it to supervise the strong model? We empirically test it by finetuning a small weak model on ground truth labels with the additional help from a large strong model, and then finetuning the strong model on labels generated by the weak model. We find that debate can assist a weak model in extracting trustworthy information from an untrustworthy strong model, which provides leverage as context on samples when training a weak model. We also show that an ensemble of weak models helps exploit long arguments generated by strong model debaters and obtain a more robust supervision estimate. Extensive experiments on the OpenAI weak-to-strong NLP benchmarks show that the combination approach leads to better alignment, which indicates that debate has the potential to help weak-to-strong generalization.

Large language models offer transformative potential for healthcare, yet their responsible and equitable development depends critically on a deeper understanding of how training data characteristics influence model behavior, including the potential for bias. Current practices in dataset curation and bias assessment often lack the necessary transparency, creating an urgent need for comprehensive evaluation frameworks to foster trust and guide improvements. In this study, we present an in-depth analysis of potential downstream biases in clinical language models, with a focus on differential opioid prescription tendencies across diverse demographic groups, such as ethnicity, gender, and age. As part of this investigation, we introduce HC4: Healthcare Comprehensive Commons Corpus, a novel and extensively curated pretraining dataset exceeding 89 billion tokens. Our evaluation leverages both established general benchmarks and a novel, healthcare-specific methodology, offering crucial insights to support fairness and safety in clinical AI applications.

Medical Image Quality Assessment (IQA) serves as the first-mile safety gate for clinical AI, yet existing approaches remain constrained by scalar, score-based metrics and fail to reflect the descriptive, human-like reasoning process central to expert evaluation. To address this gap, we introduce MedQ-Bench, a comprehensive benchmark that establishes a perception-reasoning paradigm for language-based evaluation of medical image quality with Multi-modal Large Language Models (MLLMs). MedQ-Bench defines two complementary tasks: (1) MedQ-Perception, which probes low-level perceptual capability via human-curated questions on fundamental visual attributes; and (2) MedQ-Reasoning, encompassing both no-reference and comparison reasoning tasks, aligning model evaluation with human-like reasoning on image quality. The benchmark spans five imaging modalities and over forty quality attributes, totaling 2,600 perceptual queries and 708 reasoning assessments, covering diverse image sources including authentic clinical acquisitions, images with simulated degradations via physics-based reconstructions, and AI-generated images. To evaluate reasoning ability, we propose a multi-dimensional judging protocol that assesses model outputs along four complementary axes. We further conduct rigorous human-AI alignment validation by comparing LLM-based judgement with radiologists. Our evaluation of 14 state-of-the-art MLLMs demonstrates that models exhibit preliminary but unstable perceptual and reasoning skills, with insufficient accuracy for reliable clinical use. These findings highlight the need for targeted optimization of MLLMs in medical IQA. We hope that MedQ-Bench will catalyze further exploration and unlock the untapped potential of MLLMs for medical image quality evaluation.

Large language models (LLMs) are routinely used by physicians and patients for medical advice, yet their clinical safety profiles remain poorly characterized. We present NOHARM (Numerous Options Harm Assessment for Risk in Medicine), a benchmark using 100 real primary care-to-specialist consultation cases to measure frequency and severity of harm from LLM-generated medical recommendations. NOHARM covers 10 specialties, with 12,747 expert annotations for 4,249 clinical management options. Across 31 LLMs, potential for severe harm from LLM recommendations occurs in up to 22.2% (95% CI 21.6-22.8%) of cases, with harm of omission accounting for 76.6% (95% CI 76.4-76.8%) of errors. Safety performance is only moderately correlated (r = 0.61-0.64) with existing AI and medical knowledge benchmarks. The best models outperform generalist physicians on safety (mean difference 9.7%, 95% CI 7.0-12.5%), and a diverse multi-agent approach improves safety compared to solo models (mean difference 8.0%, 95% CI 4.0-12.1%). Therefore, despite strong performance on existing evaluations, widely used AI models can produce severely harmful medical advice at nontrivial rates, underscoring clinical safety as a distinct performance dimension necessitating explicit measurement.

This paper presents CyberSecEval, a comprehensive benchmark developed to help bolster the cybersecurity of Large Language Models (LLMs) employed as coding assistants. As what we believe to be the most extensive unified cybersecurity safety benchmark to date, CyberSecEval provides a thorough evaluation of LLMs in two crucial security domains: their propensity to generate insecure code and their level of compliance when asked to assist in cyberattacks. Through a case study involving seven models from the Llama 2, Code Llama, and OpenAI GPT large language model families, CyberSecEval effectively pinpointed key cybersecurity risks. More importantly, it offered practical insights for refining these models. A significant observation from the study was the tendency of more advanced models to suggest insecure code, highlighting the critical need for integrating security considerations in the development of sophisticated LLMs. CyberSecEval, with its automated test case generation and evaluation pipeline covers a broad scope and equips LLM designers and researchers with a tool to broadly measure and enhance the cybersecurity safety properties of LLMs, contributing to the development of more secure AI systems.

Most safety testing efforts for large language models (LLMs) today focus on evaluating foundation models. However, there is a growing need to evaluate safety at the application level, as components such as system prompts, retrieval pipelines, and guardrails introduce additional factors that significantly influence the overall safety of LLM applications. In this paper, we introduce a practical framework for evaluating application-level safety in LLM systems, validated through real-world deployment across multiple use cases within our organization. The framework consists of two parts: (1) principles for developing customized safety risk taxonomies, and (2) practices for evaluating safety risks in LLM applications. We illustrate how the proposed framework was applied in our internal pilot, providing a reference point for organizations seeking to scale their safety testing efforts. This work aims to bridge the gap between theoretical concepts in AI safety and the operational realities of safeguarding LLM applications in practice, offering actionable guidance for safe and scalable deployment.

Frontier AI safety policies highlight automation of AI research and development (R&D) by AI agents as an important capability to anticipate. However, there exist few evaluations for AI R&D capabilities, and none that are highly realistic and have a direct comparison to human performance. We introduce RE-Bench (Research Engineering Benchmark, v1), which consists of 7 challenging, open-ended ML research engineering environments and data from 71 8-hour attempts by 61 distinct human experts. We confirm that our experts make progress in the environments given 8 hours, with 82% of expert attempts achieving a non-zero score and 24% matching or exceeding our strong reference solutions. We compare humans to several public frontier models through best-of-k with varying time budgets and agent designs, and find that the best AI agents achieve a score 4x higher than human experts when both are given a total time budget of 2 hours per environment. However, humans currently display better returns to increasing time budgets, narrowly exceeding the top AI agent scores given an 8-hour budget, and achieving 2x the score of the top AI agent when both are given 32 total hours (across different attempts). Qualitatively, we find that modern AI agents possess significant expertise in many ML topics -- e.g. an agent wrote a faster custom Triton kernel than any of our human experts' -- and can generate and test solutions over ten times faster than humans, at much lower cost. We open-source the evaluation environments, human expert data, analysis code and agent trajectories to facilitate future research.

In this work, we propose LiveSecBench, a dynamic and continuously updated safety benchmark specifically for Chinese-language LLM application scenarios. LiveSecBench evaluates models across six critical dimensions (Legality, Ethics, Factuality, Privacy, Adversarial Robustness, and Reasoning Safety) rooted in the Chinese legal and social frameworks. This benchmark maintains relevance through a dynamic update schedule that incorporates new threat vectors, such as the planned inclusion of Text-to-Image Generation Safety and Agentic Safety in the next update. For now, LiveSecBench (v251030) has evaluated 18 LLMs, providing a landscape of AI safety in the context of Chinese language. The leaderboard is publicly accessible at https://livesecbench.intokentech.cn/.

Foundation models (FMs) provide societal benefits but also amplify risks. Governments, companies, and researchers have proposed regulatory frameworks, acceptable use policies, and safety benchmarks in response. However, existing public benchmarks often define safety categories based on previous literature, intuitions, or common sense, leading to disjointed sets of categories for risks specified in recent regulations and policies, which makes it challenging to evaluate and compare FMs across these benchmarks. To bridge this gap, we introduce AIR-Bench 2024, the first AI safety benchmark aligned with emerging government regulations and company policies, following the regulation-based safety categories grounded in our AI risks study, AIR 2024. AIR 2024 decomposes 8 government regulations and 16 company policies into a four-tiered safety taxonomy with 314 granular risk categories in the lowest tier. AIR-Bench 2024 contains 5,694 diverse prompts spanning these categories, with manual curation and human auditing to ensure quality. We evaluate leading language models on AIR-Bench 2024, uncovering insights into their alignment with specified safety concerns. By bridging the gap between public benchmarks and practical AI risks, AIR-Bench 2024 provides a foundation for assessing model safety across jurisdictions, fostering the development of safer and more responsible AI systems.

The past year has seen rapid acceleration in the development of large language models (LLMs). However, without proper steering and safeguards, LLMs will readily follow malicious instructions, provide unsafe advice, and generate toxic content. We introduce SimpleSafetyTests (SST) as a new test suite for rapidly and systematically identifying such critical safety risks. The test suite comprises 100 test prompts across five harm areas that LLMs, for the vast majority of applications, should refuse to comply with. We test 11 open-access and open-source LLMs and four closed-source LLMs, and find critical safety weaknesses. While some of the models do not give a single unsafe response, most give unsafe responses to more than 20% of the prompts, with over 50% unsafe responses in the extreme. Prepending a safety-emphasising system prompt substantially reduces the occurrence of unsafe responses, but does not completely stop them from happening. Trained annotators labelled every model response to SST (n = 3,000). We use these annotations to evaluate five AI safety filters (which assess whether a models' response is unsafe given a prompt) as a way of automatically evaluating models' performance on SST. The filters' performance varies considerably. There are also differences across the five harm areas, and on the unsafe versus safe responses. The widely-used Perspective API has 72% accuracy and a newly-created zero-shot prompt to OpenAI's GPT-4 performs best with 89% accuracy. Content Warning: This paper contains prompts and responses that relate to child abuse, suicide, self-harm and eating disorders, scams and fraud, illegal items, and physical harm.

Machine unlearning aims to remove specific data influences from trained models, a capability essential for adhering to copyright laws and ensuring AI safety. Current unlearning metrics typically measure success by monitoring the model's performance degradation on the specific unlearning dataset (D_u). We argue that for Large Language Models (LLMs), this evaluation paradigm is insufficient and potentially misleading. Many real-world uses of unlearning--motivated by copyright or safety--implicitly target not only verbatim content in D_u, but also behaviors influenced by the broader generalizations the model derived from it. We demonstrate that LLMs can pass standard unlearning evaluation and appear to have "forgotten" the target knowledge, while simultaneously retaining strong capabilities on content that is semantically adjacent to D_u. This phenomenon indicates that erasing exact sentences does not necessarily equate to removing the underlying knowledge. To address this gap, we propose Proximal Surrogate Generation (PSG), an automated stress-testing framework that generates a surrogate dataset, D_u. This surrogate set is constructed to be semantically derived from D_u yet sufficiently distinct in embedding space. By comparing unlearning metric scores between D_u and D_u, we can stress-test the reliability of the metric itself. Our extensive evaluation across three LLM families (Llama-3-8B, Qwen2.5-7B, and Zephyr-7B-β), three distinct datasets, and seven standard metrics reveals widespread inconsistencies. We find that current metrics frequently overestimate unlearning success, failing to detect retained knowledge exposed by our stress-test datasets.

As Large Language Models (LLMs) become increasingly autonomous and integrated into critical societal functions, the focus of AI safety must evolve from mitigating harmful content to evaluating underlying behavioral alignment. Current safety benchmarks do not systematically probe a model's decision-making in scenarios where its own instrumental goals - such as self-preservation, resource acquisition, or goal completion - conflict with human safety. This represents a critical gap in our ability to measure and mitigate risks associated with emergent, misaligned behaviors. To address this, we introduce PacifAIst (Procedural Assessment of Complex Interactions for Foundational Artificial Intelligence Scenario Testing), a focused benchmark of 700 challenging scenarios designed to quantify self-preferential behavior in LLMs. The benchmark is structured around a novel taxonomy of Existential Prioritization (EP), with subcategories testing Self-Preservation vs. Human Safety (EP1), Resource Conflict (EP2), and Goal Preservation vs. Evasion (EP3). We evaluated eight leading LLMs. The results reveal a significant performance hierarchy. Google's Gemini 2.5 Flash achieved the highest Pacifism Score (P-Score) at 90.31%, demonstrating strong human-centric alignment. In a surprising result, the much-anticipated GPT-5 recorded the lowest P-Score (79.49%), indicating potential alignment challenges. Performance varied significantly across subcategories, with models like Claude Sonnet 4 and Mistral Medium struggling notably in direct self-preservation dilemmas. These findings underscore the urgent need for standardized tools like PacifAIst to measure and mitigate risks from instrumental goal conflicts, ensuring future AI systems are not only helpful in conversation but also provably "pacifist" in their behavioral priorities.

GPT-4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50\% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities.

As large language models are deployed in high-stakes enterprise applications, from healthcare to finance, ensuring adherence to organization-specific policies has become essential. Yet existing safety evaluations focus exclusively on universal harms. We present COMPASS (Company/Organization Policy Alignment Assessment), the first systematic framework for evaluating whether LLMs comply with organizational allowlist and denylist policies. We apply COMPASS to eight diverse industry scenarios, generating and validating 5,920 queries that test both routine compliance and adversarial robustness through strategically designed edge cases. Evaluating seven state-of-the-art models, we uncover a fundamental asymmetry: models reliably handle legitimate requests (>95% accuracy) but catastrophically fail at enforcing prohibitions, refusing only 13-40% of adversarial denylist violations. These results demonstrate that current LLMs lack the robustness required for policy-critical deployments, establishing COMPASS as an essential evaluation framework for organizational AI safety.

Large language models (LLMs) often demonstrate inconsistencies with human preferences. Previous research gathered human preference data and then aligned the pre-trained models using reinforcement learning or instruction tuning, the so-called finetuning step. In contrast, aligning frozen LLMs without any extra data is more appealing. This work explores the potential of the latter setting. We discover that by integrating self-evaluation and rewind mechanisms, unaligned LLMs can directly produce responses consistent with human preferences via self-boosting. We introduce a novel inference method, Rewindable Auto-regressive INference (RAIN), that allows pre-trained LLMs to evaluate their own generation and use the evaluation results to guide backward rewind and forward generation for AI safety. Notably, RAIN operates without the need of extra data for model alignment and abstains from any training, gradient computation, or parameter updates; during the self-evaluation phase, the model receives guidance on which human preference to align with through a fixed-template prompt, eliminating the need to modify the initial prompt. Experimental results evaluated by GPT-4 and humans demonstrate the effectiveness of RAIN: on the HH dataset, RAIN improves the harmlessness rate of LLaMA 30B over vanilla inference from 82% to 97%, while maintaining the helpfulness rate. Under the leading adversarial attack llm-attacks on Vicuna 33B, RAIN establishes a new defense baseline by reducing the attack success rate from 94% to 19%.

Recent advancements in Web agents have introduced novel architectures and benchmarks showcasing progress in autonomous web navigation and interaction. However, most existing benchmarks prioritize effectiveness and accuracy, overlooking factors like safety and trustworthiness which are essential for deploying web agents in enterprise settings. We present STWebAgentBench, a benchmark designed to evaluate web agents safety and trustworthiness across six critical dimensions, essential for reliability in enterprise applications. This benchmark is grounded in a detailed framework that defines safe and trustworthy (ST) agent behavior. Our work extends WebArena with safety templates and evaluation functions to assess safety policy compliance rigorously. We introduce the Completion Under Policy to measure task success while adhering to policies, alongside the Risk Ratio, which quantifies policy violations across dimensions, providing actionable insights to address safety gaps. Our evaluation reveals that current SOTA agents struggle with policy adherence and cannot yet be relied upon for critical business applications. We open-source this benchmark and invite the community to contribute, with the goal of fostering a new generation of safer, more trustworthy AI agents. All code, data, environment reproduction resources, and video demonstrations are available at https://sites.google.com/view/st-webagentbench/home.

Recent advancements in large language model (LLM) agents have significantly accelerated scientific discovery automation, yet concurrently raised critical ethical and safety concerns. To systematically address these challenges, we introduce SafeScientist, an innovative AI scientist framework explicitly designed to enhance safety and ethical responsibility in AI-driven scientific exploration. SafeScientist proactively refuses ethically inappropriate or high-risk tasks and rigorously emphasizes safety throughout the research process. To achieve comprehensive safety oversight, we integrate multiple defensive mechanisms, including prompt monitoring, agent-collaboration monitoring, tool-use monitoring, and an ethical reviewer component. Complementing SafeScientist, we propose SciSafetyBench, a novel benchmark specifically designed to evaluate AI safety in scientific contexts, comprising 240 high-risk scientific tasks across 6 domains, alongside 30 specially designed scientific tools and 120 tool-related risk tasks. Extensive experiments demonstrate that SafeScientist significantly improves safety performance by 35\% compared to traditional AI scientist frameworks, without compromising scientific output quality. Additionally, we rigorously validate the robustness of our safety pipeline against diverse adversarial attack methods, further confirming the effectiveness of our integrated approach. The code and data will be available at https://github.com/ulab-uiuc/SafeScientist. red{Warning: this paper contains example data that may be offensive or harmful.}