The EU AI Act just gained a new banned practice, and the omnibus that added it is now adopted

The digital omnibus that amends the EU AI Act is best known for pushing the high-risk deadlines back by more than a year. A detail most planning notes skip is that it also lengthened the list of practices the Act bans outright, and as of the end of June 2026 that change is adopted law rather than a proposal on the table.

From proposal to adopted law

The package moved through its final stages this summer. The European Parliament granted its final approval on 16 June 2026, and the Council of the EU formally adopted the text on 29 June 2026, the last legislative step before publication in the Official Journal and entry into force. Until that publication lands the pre-omnibus wording technically still governs, so the practical takeaway is that the direction is now settled, not that every date flips overnight. It reached this point in front of the 2 August 2026 milestone it was written to relieve.

The new line on the banned list

Article 5 already bans a short list of unacceptable-risk uses: social scoring that leads to disproportionate detrimental treatment, untargeted scraping of facial images to build recognition databases, emotion recognition at work or in education outside genuine medical or safety uses, certain biometric categorisation, and a handful of others. The omnibus adds one more. AI systems designed to generate non-consensual intimate imagery, including so-called nudification tools and material that amounts to child sexual abuse material, are now explicitly prohibited.

The timing sits apart from the rest of Article 5. The original prohibitions have applied since 2 February 2025, but this new one carries a transitional period: compliance is required by 2 December 2026. So it is a ban you can plan against rather than one that took effect the day the text was agreed.

What did not change

• The other Article 5 prohibitions still stand, and have applied since 2 February 2025.
• The high-risk deadlines still moved: stand-alone Annex III systems to 2 December 2027, and AI embedded in regulated products under Annex I to 2 August 2028.
• The Article 50 transparency duties still start on 2 August 2026, with one grace period the omnibus added: for systems already on the market before that date, the machine-readable marking obligation is deferred to 2 December 2026.

Who this actually reaches

Most businesses building ordinary AI features, a support chatbot, a summariser, a recommendation model, are nowhere near a prohibited practice, and this addition does not change that. It targets a narrow and specific category of system. But it is a reminder that the banned list is not frozen, and the first compliance question is still the same one it always was: is your system prohibited, high-risk, or only subject to transparency duties? Answer that before you budget for anything, because most systems land in the lighter tiers and owe far less than the headline fines suggest.

Our free EU AI Act risk classifier walks the Article 5, Annex III and Article 50 branches so you can see which tier, and which deadline, is yours. The companion post on the postponed high-risk deadlines covers the timeline change in full, and the guide on which tier your system is in walks the whole decision step by step.