The EU’s Artificial Intelligence Act (Regulation (EU) 2024/1689) has applied in most respects since 2 August 2026. Before that date, the EU amended it. The Digital Omnibus on AI, Regulation (EU) 2026/1744 of 8 July 2026, was agreed by Council and Parliament negotiators on 7 May, approved by Parliament’s plenary on 16 June (423 for, 57 against, 174 abstentions), adopted by the Council on 29 June, published in the Official Journal on 24 July and in force from 27 July 2026.

What applies on 2 December 2026

  • A new prohibited practice: AI systems generating non-consensual sexual or intimate content and child sexual abuse material. Prohibited practices in the AI Act carry its highest penalty band.
  • Labelling deadline for older systems: providers of AI systems that generate synthetic content, including general-purpose AI systems, that were on the market before 2 August 2026 must comply with Article 50(2) (machine-readable marking of AI-generated content) by 2 December. The omnibus cut that grace period from six months to three.

What was delayed, and what was not

  • High-risk systems listed in Annex III (including biometrics, critical infrastructure, education, employment, migration, asylum and border control): from 2 December 2027.
  • High-risk AI embedded in regulated products (Annex I, such as toys and lifts): from 2 August 2028. AI-enabled machinery was taken out of the AI Act’s direct high-risk regime and will be handled under the EU Machinery Regulation, with the Commission required to add AI safety requirements there by delegated act.
  • Already in force: the original bans (since February 2025), obligations for general-purpose AI model providers (since August 2025), and, since 2 August 2026, the Article 50 transparency rules and national and EU enforcement.
  • Negotiators reinstated two things the Commission had proposed to drop: registration in the EU database for providers who consider their system exempt from the high-risk class, and the “strict necessity” test for processing sensitive personal data to detect bias.
“Prohibitions, for AI systems generating non-consensual sexual deepfakes and child sexual abuse material, start to apply.” — European Commission AI Act Service Desk timeline, entry for 2 December 2026
Citizen impact: For people in the EU, the nearest change is protective: from 2 December, offering AI that produces sexual deepfakes of real people without consent, or child abuse material, is a banned practice under EU law, on top of national criminal laws. The trade-off is time: the rules meant to govern AI used to screen job applicants, grade students or check identities at borders now start 16 months later than the original August 2026 date. Who gains from the delay: developers and deployers of high-risk systems. Who waits: people those systems assess.

What this piece refuses

Do not write that the EU “scrapped” or “paused” the AI Act. The omnibus delays one block of obligations and adds a ban; the rest of the Act applies. We have not reported fine amounts for the new ban, because enforcement sits with national authorities and the AI Office and none has been levied.

Bottom line

Verified from the Commission’s AI Act pages, the Parliament legislative record and the EUR-Lex summary: Regulation (EU) 2026/1744 has been in force since 27 July 2026. The ban on AI-generated non-consensual intimate content and CSAM applies from 2 December 2026; high-risk rules start 2 December 2027 and 2 August 2028.