EU delays its high-risk AI rules to late 2027 — but not everything is postponed
The EU's 'AI Omnibus' package — given final approval by the Council on June 29 — pushes the AI Act's high-risk obligations from August 2, 2026 to December 2, 2027. Not everything moves: bans on the worst practices and machine-readable labeling of AI-generated content keep their own deadlines (December 2026), and national enforcement bodies are already operational. (Correction: an earlier version of this story said the high-risk rules would take effect on August 2, 2026.)
Why it mattersCompanies get sixteen extra months for the heaviest compliance work — but this isn't a full pause. Content-labeling duties and outright bans stay on schedule, and Europe's enforcement machinery is already running. The template for global AI regulation is still being set; just more slowly than planned.
▶ Related video: The EU AI Act Rules Everyone Thinks Got Delayed — and the Ones Taking Effect Anyway
Read in the app — free, in 9 languages
Related stories
Nevada just cleared up to 8,000 driverless taxis for Las Vegas — Tesla's ceiling went from 10 to 5,000 in five weeks2026-08-21OpenAI has started a team whose stated worry is that a company like OpenAI ends up with too much power
2026-08-21OpenAI says it can police misuse without keeping your company's data — and it is pointing at Anthropic while saying so
2026-08-20A Florida officer ran his ex-wife's plate 717 times. A Kansas city has just switched its cameras off.
2026-08-20Six months ago Hollywood sent ByteDance a cease-and-desist. This week they signed a truce — and no money changed hands.
2026-08-20