EU AI Act: what actually applies in late 2026 after high-risk rules were delayed

Contents
The European Union’s artificial intelligence law, the AI Act, entered into force on August 1, 2024, and has applied generally since August 2, 2026. But its timetable changed along the way: the “Digital Omnibus on AI,” given final approval by EU governments on June 29, 2026, and in force since July 27, postponed the obligations for high-risk AI. Here is what applies as of November 17, 2026, and what comes next.
What was delayed, and why?
Proposed by the European Commission on November 19, 2025, the Omnibus is part of the EU’s simplification drive. A political agreement was reached on May 7, 2026. High-risk rules originally due on August 2, 2026, now apply on December 2, 2027 for Annex III uses (hiring, education, credit, biometrics, migration, justice) and on August 2, 2028 for AI built into regulated products such as toys or lifts.
Other changes: the AI literacy duty is softened (companies must support staff training rather than guarantee a set level), relief measures for SMEs now extend to small mid-caps, and the EU AI Office gains stronger powers over systems built on general-purpose models.
| Date | What applies |
|---|---|
| August 1, 2024 | Regulation enters into force |
| February 2, 2025 | Eight prohibited practices; AI literacy duty |
| August 2, 2025 | Rules for general-purpose AI models; governance |
| August 2, 2026 | General application, enforcement by the AI Office and national authorities; transparency rules |
| December 2, 2026 | Ninth prohibition (non-consensual intimate content, child abuse material); labeling for systems already on the market |
| August 2, 2027 | Deadline for national regulatory sandboxes |
| December 2, 2027 | Annex III high-risk systems (employment, education, credit, biometrics…) |
| August 2, 2028 | High-risk AI embedded in products (Annex I: toys, lifts, medical devices…) |
What changes on December 2, 2026?
The Omnibus adds a ninth prohibited practice: systems that generate non-consensual sexual or intimate content or child sexual abuse material, including photo “nudification” apps. The ban applies from December 2, 2026.
The same day, the grace period for labeling AI-generated content ends: systems placed on the market before August 2, 2026, must make their output identifiable. That grace period was cut from six months to three. Until then, our guide to spotting deepfakes can help.
What is already banned?
Since February 2, 2025, eight practices have been prohibited in the EU. According to the Commission, they are:
- manipulative or deceptive AI that causes harm, and exploiting people’s vulnerabilities;
- social scoring;
- assessing or predicting an individual’s risk of committing a crime;
- untargeted scraping of internet or CCTV images to build facial recognition databases;
- emotion recognition in workplaces and schools;
- biometric categorization to infer protected characteristics;
- real-time remote biometric identification by police in public spaces.
What does it mean for people and businesses?
For the public, the most visible effect is transparency: a chatbot must disclose that it is AI, and deepfakes must be labeled. High-risk rules, such as those covering automated CV screening, only arrive in December 2027; we explore the stakes in AI and jobs. Education is also a high-risk area, as our piece on AI in schools explains.
For businesses, maximum fines under Article 99 are unchanged: up to €35 million or 7% of global annual turnover for prohibited practices, €15 million or 3% for other obligations, and €7.5 million or 1% for misleading information. For SMEs, the lower of the two amounts applies.
Developers of general-purpose models can also rely on the GPAI Code of Practice, which the Commission describes as a voluntary compliance tool drafted by independent experts, with guidance on transparency, copyright, and safety and security. Since August 2, 2026, the AI Office can also issue fines for non-compliance.
Frequently asked questions
Has the AI Act been suspended?
No. It has applied generally since August 2, 2026. Only the high-risk obligations were pushed back, to late 2027 and 2028.
Does it cover ChatGPT or Gemini?
Yes. General-purpose AI models have had obligations since August 2, 2025, and transparency rules apply to chatbots.
What happens on December 2, 2026?
Apps generating non-consensual intimate images or child abuse material are banned, and output from systems already on the market must be labeled as AI-generated.
What are the penalties?
Up to €35 million or 7% of worldwide turnover for the most serious breaches, under Article 99 of the regulation.
Sources
- Commission européenne – AI Act, regulatory framework for AI (page mise à jour) (October 19, 2026)
- Conseil de l’UE (relayé par le SEAE) – Artificial intelligence: Council gives final green light to simplify and streamline rules (June 30, 2026)
- Gibson Dunn – EU AI Act Omnibus Agreement: postponed high-risk deadlines and other key changes (May 27, 2026)
- Garrigues – AI Digital Omnibus regulation has been published, redefining deadlines (July 24, 2026)
- Règlement (UE) 2024/1689, article 99 (sanctions) – texte consolidé, artificialintelligenceact.eu (July 12, 2024)
- España Digital – Aprobado el estatuto de la Agencia Española de Supervisión de la Inteligencia Artificial (August 22, 2023)




