Did the Digital Omnibus delay Article 50?
No. The EU's "Digital Omnibus" simplification package pushed back the AI Act's high-risk system rules (Annex III) from August 2026 to December 2, 2027. It did not touch Article 50. Chatbot disclosure, synthetic content marking, and deepfake labeling are all enforceable now, and have been since August 2, 2026 — along with the fines.
Why the confusion
The Digital Omnibus reached provisional agreement on May 7, 2026, as a broad simplification package aimed at easing implementation burden across several EU digital regulations, the AI Act included. Coverage of that deal — reasonably — focused on the biggest relief it granted: a roughly 16-month deferral of the high-risk AI system regime under Annex III, the part of the Act most enterprises were racing to certify against. That headline got generalized in a lot of secondhand summaries into "the AI Act got delayed," full stop. It didn't.
What actually moved
- Annex III high-risk system obligations — certification, conformity assessment, the heaviest compliance lift in the Act — moved from August 2, 2026 to December 2, 2027.
What did not move
- Article 50(1) — AI-interaction disclosure (chatbots, voice assistants) — in force since August 2, 2026.
- Article 50(4) — deepfake and AI-generated public-interest text disclosure — in force since August 2, 2026.
- Enforcement and fines — national market surveillance authorities gained the power to fine up to €15 million or 3% of global turnover for Article 50 breaches, effective the same date. This power was not deferred.
There's exactly one genuine grace period inside Article 50 itself, and it's narrow: systems generating synthetic content that were already on the market before August 2, 2026 get until December 2, 2026 to implement the Article 50(2) machine-readable marking requirement. Any system launched on or after August 2, 2026 gets no grace period at all — it has to mark from day one.
The practical takeaway
If your business has a chatbot, or generates AI images/audio/video/text, "the AI Act got pushed back" is not a safe assumption to operate on. The specific rules most small and mid-size teams actually run into — disclose the chatbot, mark the synthetic content, label the deepfake — are live, enforceable, and carry real fine exposure right now. The part that got real relief is the heavier, enterprise-facing high-risk certification regime, which was never the piece most SaaS products or content tools were dealing with in the first place.
Not sure where you stand? Run the free scanner to check whether your site discloses AI chat interaction and marks synthetic content correctly — or work through the full checklist.
Run the free scan → See the full checklist →