The EU’s risk-based law for AI systems. Its bans and general-purpose AI rules already apply; transparency duties apply from 2 August 2026; new bans on AI “nudifier” imagery apply from 2 December 2026; high-risk rules were moved to December 2027 and August 2028.
The EU AI Act sorts AI systems by risk. A few practices are banned outright, high-risk uses such as hiring, credit and education face strict duties, and anyone offering chatbots or synthetic media has to be transparent about it.
In July 2026 the Digital Omnibus on AI (Regulation (EU) 2026/1744) moved the high-risk deadlines back and added new bans on AI-generated non-consensual intimate imagery and child sexual abuse material.
Key dates
The AI Act enters into force
Bans on prohibited AI practices apply
Obligations for general-purpose AI models apply
Transparency duties apply: disclose AI chatbots, label deepfakes
New bans on AI nudifier imagery and child sexual abuse material apply
High-risk rules for stand-alone systems (Annex III) apply
High-risk rules for AI in regulated products (Annex I) apply
Digital Omnibus on AI (Regulation (EU) 2026/1744) entered into force; high-risk dates moved to 2 Dec 2027 and 2 Aug 2028; new bans added from 2 Dec 2026