The AI Act Timeline
Adopted in March 2024, the European AI Act comes into force in phases:
- February 2025: bans on unacceptable risk systems
- August 2025: obligations for general-purpose models (GPAI)
- August 2026: full obligations for high-risk systems
We're in April 2026. Most provisions are now applicable.
The Bans: What Is Now Illegal
Subliminal Manipulation
AI systems designed to manipulate behavior subconsciously are banned.
Social Scoring
No Chinese-style social credit system in Europe.
Mass Facial Recognition
Real-time biometric recognition in public spaces is banned, with strict exceptions.
High-Risk Systems: Enhanced Obligations
The AI Act defines "high-risk sectors" where AI systems must meet strict obligations: recruitment, education, essential services, justice, migration, critical infrastructure.
Obligations include: conformity assessment, technical documentation, transparency, human oversight, data management, robustness.
General Purpose AI Models
Foundation models like GPT-4, Claude, or Mistral have specific obligations: technical documentation, copyright policy, training data summary.
Sanctions: Deterrent Fines
- Prohibited practices: up to €35 million or 7% of global turnover
- Non-compliance: up to €15 million or 3% of global turnover
The Brussels Effect
Like GDPR, the AI Act becomes a de facto global standard. Companies serving the European market must comply.
The AI Act isn't a constraint, it's the new rules of the game. Better to master it.