The Artificial Intelligence Act received its first significant revision on June 29, when the Council of the EU formally approved the Digital Omnibus on AI — a legislative package that adjusts compliance timelines and introduces new prohibitions.
Deadlines pushed back
The most consequential change affects stand-alone Annex III high-risk AI systems — software applied in employment screening, credit scoring, education, and public-service delivery. Operators faced a compliance deadline of August 2, 2026; that date now moves to December 2, 2027, a 16-month extension.
For high-risk AI systems embedded in safety-regulated products — including medical devices, toys, and industrial machinery covered under Annex I — the deadline shifts further to August 2, 2028.
EU negotiators cited incomplete technical standards as the practical reason for the delay. “By providing greater legal certainty and ensuring more harmonised implementation of AI rules across the EU, we are creating conditions for innovation and growth,” said Marilena Raouna, Cyprus’s Deputy Minister for European Affairs, following the Council vote.
New bans, arriving sooner
While compliance deadlines moved back, new prohibitions come into force earlier. From December 2026, EU law will explicitly forbid AI systems that generate non-consensual sexual imagery of real people, digitally remove clothing from images, or produce child sexual abuse material (CSAM).
What stays on schedule
Transparency obligations — requiring disclosure when content such as text, images, or audio is AI-generated — remain broadly on their original timeline. Providers must implement these solutions by December 2, 2026, a window the Digital Omnibus actually tightened compared with the earlier six-month grace period.
The regulation enters into force three days after publication in the EU’s Official Journal, expected shortly.