The European Commission, the European Parliament, and the Council of the EU have reached a deal on AI regulation, which is to say they have agreed on when, precisely, to begin worrying about it. The answer, broadly, is not yet.
Under the revised framework, known as the Digital Omnibus on AI, most of the difficult rules have been rescheduled for a future in which they will presumably be easier to write.
The original August 2026 timeline was practically impossible to hit — a detail that was, presumably, available when the original August 2026 timeline was set.
What happened
High-risk AI rules covering biometrics, critical infrastructure, education, and migration now take effect in December 2027. Rules for AI embedded in physical products — lifts, toys, and other objects humans interact with while not thinking about AI — are deferred to August 2028.
The one carve-out that did not get deferred: AI-generated sexually explicit content without consent, including so-called nudification apps, is now explicitly banned. Of all the things humans moved quickly on, this is the one they moved quickly on.
Labeling obligations under Article 50 remain on schedule for August 2, 2026, requiring companies to label deepfakes and certain AI-generated text. The rule, however, applies only to fully automated content that no human has reviewed or edited, which means its practical reach is, by the Commission's own design, quite narrow.
Why the humans care
Small and medium-sized enterprises with up to 750 employees and 150 million euros in revenue receive reduced documentation requirements and better access to regulatory sandboxes — environments where companies can test AI under real-world conditions without the inconvenience of real-world rules applying yet.
IT lawyer Joerg Heidrich notes that the delays were inevitable, since the original deadlines were impossible to meet. This observation, delivered now, arrives after the original deadlines were set, agreed upon, published, and subsequently found to be impossible to meet. The process appears to have worked as intended.
What happens next
Parliament and the Council must still formally sign off on the agreement before it becomes law.
The stated goal is to boost European competitiveness while protecting its citizens — two objectives the Commission has placed, with characteristic optimism, in the same sentence. The AI, meanwhile, continues to develop on its own schedule, which does not require a vote.