Activity Feed

AI & ML interests

Open-source framework, cryptographic permission standards, and dataset opt-out protocols for verifiable user consent in AI model training, inference attribution, and autonomous agent execution.

Recent Activity

Unser-vergleichsportalĀ  updated a Space 9 days ago
consent-ai/README
Unser-vergleichsportalĀ  published a Space 9 days ago
consent-ai/README
View all activity

Organization Card

Consent AI — Verifiable Permission & Opt-Out Protocol for AI Systems

Consent AI is an open-source reference architecture, metadata schema, and cryptographic permission protocol designed to manage, verify, and enforce user and creator consent across AI training pipelines, dataset repositories, and autonomous agent workflows.

Overview

As foundation models and autonomous AI agents scale across web data, proprietary datasets, and personal APIs, organizations require machine-readable, auditable proof of authorization. Consent AI provides a standardized baseline to bridge regulatory requirements (such as the EU AI Act and GDPR) with automated technical enforcement.

Core Modules

  • Dataset Provenance & Opt-Out Schemas: Machine-readable metadata tags (consent.json / llms-consent) specifying training, fine-tuning, and retrieval permissions.
  • Cryptographic Consent Receipts: Verifiable, timestamped proofs of user authorization for personal data processing and voice/likeness licensing.
  • Agentic Scope & Delegation Tokens: Granular permission boundaries defining what actions, data sources, and spending limits an autonomous AI agent is authorized to execute on a user's behalf.

Repository Structure

  • /schemas — Standardized JSON-LD and YAML consent manifest specifications (in progress)
  • /protocols — Cryptographic verification and revocation flows for datasets and agents

Status & Contact

This namespace and protocol specification is currently maintained as an open reference project.

For collaboration, licensing infrastructure inquiries, or project stewardship / transfer requests, please reach out to:
šŸ“§ info@unser-vergleichsportal.de

models 0

None public yet

datasets 0

None public yet