| { |
| "title": "Cross-Regulatory Harmonization Framework: GDPR, EU AI Act & NIS2 Directive", |
| "version": "1.0.0", |
| "regulations": { |
| "GDPR": { |
| "official_title": "Regulation (EU) 2016/679 (General Data Protection Regulation)", |
| "celex": "32016R0679", |
| "scope": "Protection of natural persons with regard to personal data processing." |
| }, |
| "EU_AI_ACT": { |
| "official_title": "Regulation (EU) 2024/1689 (Artificial Intelligence Act)", |
| "celex": "32024R1689", |
| "scope": "Harmonized rules on the placing on the market, putting into service, and use of artificial intelligence systems in the Union." |
| }, |
| "NIS2": { |
| "official_title": "Directive (EU) 2022/2555 (Network and Information Systems Directive 2)", |
| "celex": "32022L2555", |
| "scope": "Measures for a high common level of cybersecurity across essential and important entities across the Union." |
| } |
| }, |
| "cross_regulatory_rules": [ |
| { |
| "id": "CROSS-001", |
| "title": "Prohibited Biometric Categorization & Sensitive Attributes", |
| "domain": "Artificial Intelligence & Special Category Data", |
| "primary_articles": { |
| "gdpr": "Art. 9(1), Art. 4(14)", |
| "ai_act": "Art. 5(1)(c), Art. 5(1)(d)", |
| "nis2": "N/A" |
| }, |
| "severity": "CRITICAL_PROHIBITION", |
| "deontic_rule": "PROHIBITION", |
| "description": "AI systems that deduce or infer sensitive personal attributes (race, political opinions, trade union membership, religious beliefs, sexual orientation) from biometric data are strictly prohibited under AI Act Art. 5(1)(c) and constitute unlawful processing under GDPR Art. 9(1).", |
| "exceptions": "Lawfully acquired biometric data in criminal law enforcement subject to judicial warrant." |
| }, |
| { |
| "id": "CROSS-002", |
| "title": "Dual Incident Notification SLA (24h Early Warning + 72h DPA)", |
| "domain": "Cybersecurity & Breach Escalation", |
| "primary_articles": { |
| "gdpr": "Art. 33(1), Art. 34(1)", |
| "ai_act": "Art. 73 (Serious Incident Reporting)", |
| "nis2": "Art. 23(1), Art. 23(4)" |
| }, |
| "severity": "CRITICAL_TIMELINE", |
| "deontic_rule": "OBLIGATION", |
| "description": "When a cybersecurity incident impacts both network continuity and personal data confidentiality, dual notifications are mandatory: (1) NIS2 Early Warning to CSIRT within 24 hours, (2) GDPR DPA notification within 72 hours, and (3) NIS2 Detailed Incident Notification within 72 hours.", |
| "statutory_deadlines": { |
| "nis2_early_warning_hours": 24, |
| "gdpr_dpa_notification_hours": 72, |
| "nis2_final_report_days": 30 |
| } |
| }, |
| { |
| "id": "CROSS-003", |
| "title": "High-Risk AI Data Governance & Accuracy Mandate", |
| "domain": "Data Quality & Machine Learning Governance", |
| "primary_articles": { |
| "gdpr": "Art. 5(1)(d) (Accuracy), Art. 25 (Privacy by Design)", |
| "ai_act": "Art. 10 (Data and Data Governance)", |
| "nis2": "N/A" |
| }, |
| "severity": "HIGH_GOVERNANCE", |
| "deontic_rule": "OBLIGATION", |
| "description": "Training, validation, and testing datasets for High-Risk AI systems must be relevant, representative, and to the best extent possible free of errors. Controllers must apply data minimization and privacy-preserving techniques (synthetic data, differential privacy) during model training.", |
| "mandatory_measures": [ |
| "Statistical bias auditing", |
| "Documented data provenance & curation", |
| "Data minimization in training pipelines" |
| ] |
| }, |
| { |
| "id": "CROSS-004", |
| "title": "Automated Profiling, Credit Scoring & Essential Services Access", |
| "domain": "Automated Decision-Making & Human Oversight", |
| "primary_articles": { |
| "gdpr": "Art. 22(1), Art. 35(3)(a)", |
| "ai_act": "Annex III(5)(b), Art. 14 (Human Oversight)", |
| "nis2": "N/A" |
| }, |
| "severity": "CRITICAL_OVERSIGHT", |
| "deontic_rule": "OBLIGATION", |
| "description": "AI systems used to evaluate creditworthiness or risk in health/life insurance are classified as High-Risk AI (Annex III(5)(b)). Under GDPR Art. 22, individuals retain the right not to be subject to purely automated decisions; Article 14 AI Act mandates human oversight controls capable of overriding model outputs.", |
| "mandatory_measures": [ |
| "Mandatory DPIA prior to deployment (GDPR Art. 35)", |
| "Human-in-the-loop override mechanism (AI Act Art. 14)", |
| "Meaningful transparency on logic involved (GDPR Art. 13(2)(f))" |
| ] |
| }, |
| { |
| "id": "CROSS-005", |
| "title": "Cybersecurity Risk Management & Technical Security Measures", |
| "domain": "Technical & Organizational Measures (TOMs)", |
| "primary_articles": { |
| "gdpr": "Art. 32(1) (Security of Processing)", |
| "ai_act": "Art. 15 (Accuracy, Robustness & Cybersecurity)", |
| "nis2": "Art. 21 (Cybersecurity Risk-Management Measures)" |
| }, |
| "severity": "BASELINE_SECURITY", |
| "deontic_rule": "OBLIGATION", |
| "description": "Essential entities under NIS2 and controllers under GDPR must implement state-of-the-art security: multi-factor authentication (MFA), end-to-end encryption, supply chain security assessments, and resilience against adversarial AI attacks (data poisoning, model evasion).", |
| "mandatory_measures": [ |
| "MFA across all administrative access", |
| "Supply chain third-party vendor audits (NIS2 Art. 21(2)(d))", |
| "Adversarial robustness testing against model extraction" |
| ] |
| } |
| ] |
| } |
|
|
