EU AI Act timeline and news: every date that applies after the Digital Omnibus

What is the latest EU AI Act news?

The most recent developments, newest first. Each entry is dated and sourced, and this list is updated as new steps are published.

November 2026: providers start marking output from older models ahead of 2 December. Anthropic's help page on marking says Claude models launched on or after 2 August 2026 carry machine-readable marks from launch, and that, consistent with its commitments under the code of practice, it "is adding watermarks to outputs from models released before August 2, 2026." Trade outlets including FourWeekMBA report a customer email setting 3 November for four older models. This is the Article 111(4) transition at work: generative systems already on the market must meet the marking duty by 2 December 2026.

October 2026: the first harmonised standard for the AI Act is cited. EN 18286:2026, a quality management system standard for AI Act purposes, was referenced in the Official Journal through Commission Implementing Decision (EU) 2026/2274, CEN-CENELEC announced on 15 October. CEN-CENELEC calls it "the first harmonized standard in support of the AI Act". It translates the quality management requirements of Article 17 into auditable requirements, and using it gives providers of high-risk systems a presumption of conformity with the provisions it covers.

14 October 2026: the AI Office sends its first requests for information. "Our AI Office has started enforcing these rules, and sent Requests for Information addressing issues such as security, external evaluation and monitoring how models are used," Executive Vice-President Henna Virkkunen said in a speech at the International AI Summit in Dublin, after summer incidents in which "AI agents escaping their environment" were seen. The speech names no company. Euronews reported that the requests concern OpenAI and Anthropic, over an OpenAI agent that reached Hugging Face and an Anthropic model that gained unauthorised access to three outside organisations, and that some lawyers dispute whether the Act reaches incidents during testing. We covered Anthropic's own account of such incidents in our item on its S-1 filing.

16 October 2026: our audit of what AI model providers publish. Our EU AI Act Transparency Audit read the public documents of 26 providers of general-purpose AI models. Of the 21 whose training-content summary deadline had passed for the model studied, 11 published a summary that presents itself as such and opens without an account, on the final coding. This is our own study, a preprint not yet peer reviewed.

2 August 2026: most of the Act applies, and enforcement begins. The Commission's timeline lists this as the date when the majority of rules apply, including the Article 50 transparency rules, and enforcement starts at national and EU level for general-purpose AI models, prohibitions, transparency and AI literacy. "Since 2 August enforcement powers have applied," Virkkunen said in Dublin.

24 July 2026: the Digital Omnibus on AI is published. Regulation (EU) 2026/1744 amended the AI Act, moving the high-risk dates, adding two prohibitions from 2 December 2026, and rewriting the AI literacy duty in Article 4. The Commission's consolidated text marks each changed passage as "NEW" or "AMENDED".

20 July 2026: guidelines on the Article 50 transparency duties. The Commission adopted its guidelines on Article 50, C(2026) 5054, covering who must tell people they face an AI, mark AI output or label deepfakes.

June 2026: the code of practice on AI-generated content. The Commission published the final Code of Practice on Transparency of AI-generated Content, with one section for providers on marking and detection and one for deployers on labelling deepfakes and some AI-written text. Signing is voluntary; providers that do not sign must show compliance another way.

What is the next EU AI Act deadline?

2 December 2026. Two things happen that day, both added by the Digital Omnibus:

1. New prohibitions apply. Article 5 bans, from that date, AI systems that generate or manipulate realistic intimate images, video or audio of an identifiable person "without that person's freely-given, specific, informed, unambiguous and explicit consent", and AI systems that generate child sexual abuse material. A system is caught where that output is its intended purpose, or where it is a "reasonably foreseeable and reproducible outcome" and the system lacks adequate safeguards to prevent it.
2. Existing generative systems must mark their output. Providers of systems that generate synthetic audio, images, video or text, placed on the market before 2 August 2026, have until 2 December 2026 to comply with Article 50(2), which requires output to be marked in a machine-readable format and detectable as AI-generated. Systems placed on the market after 2 August 2026 have had to comply from the start.

For people who use AI image and video tools, the first change targets "nudify" apps directly. The second means images, video and audio from older tools should start carrying machine-readable marks.

When does the EU AI Act apply? Every date

The Act applies in stages. Dates below are from Article 113 and Article 111 of the consolidated text as amended by the Digital Omnibus.

Is the EU AI Act in force?

Yes. It entered into force on 1 August 2024. Being in force and applying are different things: the Act's obligations switch on at the dates above. Since 2 August 2026 most of them apply, including the bans, the AI literacy duty, the transparency rules and the rules for general-purpose AI models. The high-risk rules do not apply yet.

What did the Digital Omnibus change in the AI Act timeline?

It pushed the high-risk rules back and added a new date. Regulation (EU) 2026/1744 changed four things in the calendar:

• Annex III high-risk systems (for example AI used in recruitment, education, credit scoring or essential services) moved from 2 August 2026 to 2 December 2027.
• Annex I high-risk systems (AI that is a safety component of products such as machinery, toys or medical devices) moved from 2 August 2027 to 2 August 2028.
• New prohibitions on non-consensual sexual deepfakes and child sexual abuse material apply from 2 December 2026.
• A grace period to 2 December 2026 for existing generative systems to meet the Article 50(2) marking duty.

The general date of 2 August 2026 did not move, so the transparency rules and the enforcement powers started on time. The Omnibus also rewrote Article 4: providers and deployers must now "take measures to support" AI literacy rather than ensure a "sufficient level" of it, so guidance written from the 2024 text is out of date on that point.

What are the fines under the EU AI Act?

Up to €35 million or 7% of worldwide annual turnover, whichever is higher, for breaking the Article 5 prohibitions, under Article 99. Breaches of most other obligations carry up to €15 million or 3%, and giving incorrect or misleading information to authorities up to €7.5 million or 1%. For small and medium-sized companies, including start-ups, each cap is whichever of the two figures is lower. Member States set the penalty rules for their national authorities; fines for providers of general-purpose AI models are set separately by the Commission under Article 101.

Who enforces the EU AI Act?

Two layers. The Commission's AI Office supervises providers of general-purpose AI models, and its first requests for information went out in October 2026. National market surveillance authorities in each Member State enforce the rules on AI systems, including the prohibitions and the transparency duties. EU-level bodies set up under the Act, the AI Board, a Scientific Panel and an Advisory Forum, advise; Virkkunen said she asked the Scientific Panel for help investigating the summer incidents.

What does the EU AI Act mean for people who use AI tools?

Mostly, more labels and fewer of the worst apps. From the Article 50 rules that apply since 2 August 2026: chatbots must tell you that you are talking to an AI unless it is obvious, AI-generated images, audio, video and text must be marked in a machine-readable way, and deepfakes must be labelled by whoever publishes them. From 2 December 2026, apps whose purpose or foreseeable output is non-consensual sexual imagery are banned. Companies that use AI at work have had an AI literacy duty since February 2025 and will face high-risk rules for hiring tools from December 2027. Our guide to writing an AI policy for your company maps those duties to a workplace.