EU AI Act Deadline Tracker

Article 5, Banned AI systems and practices are now enforceable.

• Social scoring systems for public authorities
• Real-time remote biometric identification (with exceptions)
• Emotion recognition in workplaces and education
• AI exploiting vulnerabilities of specific groups
• Subliminal manipulation causing harm

Chapter V, General-purpose AI model providers must comply.

• Technical documentation and training data summaries
• Copyright law compliance obligations
• Systemic risk models: adversarial testing, incident reporting
• Model cards and capability evaluations

Article 50 transparency obligations have applied since 2 August 2026.

• Inform natural persons when they interact with AI
• Mark synthetic audio, image, video, or text content
• Applies to providers and deployers under Article 50 conditions

Chapter III, standalone high-risk requirements (Annex III). Applies from **2 December 2027**. The date is unconditional under Regulation (EU) 2026/1744.

• Risk management system (Article 9)
• Data governance and quality requirements (Article 10)
• Technical documentation, Annex IV (Article 11)
• Record-keeping and logging (Article 12)
• Transparency and user information (Article 13)
• Human oversight measures (Article 14)
• Accuracy, robustness, and cybersecurity (Article 15)
• Fundamental Rights Impact Assessment (Article 27)

All remaining provisions of the AI Act are enforceable. Full penalty regime active.

• Fines up to €35M or 7% of global annual turnover
• National market surveillance authorities fully operational
• EU AI Office enforcement actions
• Conformity assessments required for all high-risk systems
• Post-market monitoring obligations