The AI Omnibus entered into force across the EU on July 27, 2026, six days before Article 50’s transparency obligations become enforceable on August 2. That six-day gap is the whole story: Brussels has split the AI Act’s compliance calendar in two, keeping the consumer-facing transparency regime on schedule while giving industry a substantially longer runway on the heavy machinery.

The Council signed off on the “Omnibus VII” package on June 29. Its most consequential move pushes high-risk obligations for stand-alone Annex III systems from August 2, 2026 to December 2, 2027, a 16-month slip. High-risk systems embedded in products get until August 2, 2028. National regulatory sandboxes, originally due in 2026, are postponed one year to August 2, 2027.

What didn’t move matters just as much. Article 50 lands on time. Chatbot disclosures, deepfake labeling, synthetic-content marking, and emotion or biometric categorization notices all become enforceable August 2, with fines reaching €15 million or 3% of worldwide annual turnover. The Commission adopted its 51-page guidelines on July 20, and national market surveillance authorities are expected to apply them alongside the Code of Practice on Transparency of AI-Generated Content, which Brussels has assessed as adequate. Providers of generative AI already on the market have until December 2, 2026 to implement machine-readable marking under Article 50(2). Deepfakes generated before August 2 don’t require retroactive labeling.

The Omnibus also adds a new Article, effective December 2, 2026, prohibiting AI generation of non-consensual sexual imagery and CSAM outright.

The structural read is that Brussels has learned to triage. The politically legible parts of the Act, labels, chatbot warnings, deepfake bans, arrive on time. The technically demanding conformity assessments slide quietly to 2027 and 2028. It’s the same regulation, resequenced for what the institution can actually enforce.

Sources

Sources