Prohibited AI practices are enforceable now — here is what to check
Article 5 prohibitions took effect February 2, 2025. Six categories of AI practice are banned outright, with fines up to €35 million or 7% of global turnover.
The six categories already in force
Subliminal manipulation causing harm, exploitation of vulnerabilities (age, disability, social/economic situation), social scoring by public authorities or on their behalf, individual criminal risk prediction based solely on profiling, untargeted facial-image scraping, and emotion recognition in workplaces and education — with narrow exceptions for safety and medical use.
Real-time remote biometric identification in public spaces by law enforcement is also prohibited except under strict judicial authorization for specific serious offences.
Why early action matters
These prohibitions carry the highest penalty tier: up to €35 million or 7% of worldwide annual turnover, whichever is higher. SMEs face proportionally adjusted but still significant fines.
The February 2025 guidelines published by the European Commission provide detailed scenarios and boundary cases. Organizations should cross-reference their AI inventory against each prohibition category now.
Operational steps
Conduct a prohibition screening for every AI system in your inventory. Document the screening result and reasoning. For any system that approaches a prohibition boundary, obtain legal review and maintain that analysis as evidence.
ActLoom's classification engine flags potential prohibition overlaps during initial system registration, giving teams a head start on this critical assessment.