2 December 2026 is the next hard date in the EU AI Act. Two things happen on that day: the extra time for older generative AI systems to mark their output runs out, and two new bans take effect. This guide explains both, who they reach, and what to look at before the date.
Why this date exists
The Digital Omnibus on AI, Regulation (EU) 2026/1744, was published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026. It changed several dates in the AI Act. The best-known change moved the high-risk rules to December 2027 and August 2028. The same law also created the two 2 December 2026 dates covered here.
Change 1: marking AI-generated content (Article 50(2))
Article 50(2) of the AI Act says providers of AI systems that generate synthetic audio, images, video or text must mark the output in a machine-readable format, so that it can be detected as artificially generated or manipulated.
Who it applies to: the provider of the generative AI system, meaning the company that builds the system or puts it on the market under its own name. A business that only uses someone else's tool is a deployer and has other duties (see below).
What the date means:
- A generative system placed on the EU market before 2 August 2026 has until 2 December 2026 to comply with the marking duty.
- A system placed on the market on or after 2 August 2026 has no grace period. It had to comply from the day it was placed on the market.
The grace period is a transitional rule for existing systems. It does not move the Article 50 rules themselves, which have applied since 2 August 2026.
Change 2: two new bans (Article 5)
From 2 December 2026, the list of prohibited practices in Article 5 grows by two items. They cover AI used to create:
- non-consensual intimate content, including so-called nudifier apps, and
- child sexual abuse material.
The duties differ by role. Providers may not place on the market a system that is designed for these purposes, or that can reasonably and foreseeably be used for them without adequate technical safeguards. Deployers may not use an AI system to create or manipulate this kind of material.
The safeguard wording matters. A general-purpose image or video generator is not automatically banned. A provider that can show effective technical measures against these outputs is in a much stronger position than one that cannot. Read the exact text of the new Article 5 provisions before relying on this summary.
What already applies to you since 2 August 2026
Several transparency duties were not delayed and already apply:
- Telling people they are talking to AI (Article 50(1)). If a person interacts with an AI system, such as a chatbot, they must be told, unless it is obvious from the context.
- Labelling deepfakes (Article 50(4)). A deployer that publishes image, audio or video content that is AI-generated or manipulated and realistically resembles real people, places or events must disclose that clearly.
- AI-generated text on matters of public interest (Article 50(4)). Disclosure is required, unless the text was reviewed by a human or is under editorial control and someone takes editorial responsibility for publishing it.
The Commission adopted guidelines on these transparency duties on 20 July 2026. They are worth reading, because they explain how the Commission reads terms such as "deepfake" and cover AI agents.
Who should pay attention before 2 December
- You build or sell a generative AI tool used in the EU (text, image, audio or video) and it was on the market before 2 August 2026. Your marking deadline is 2 December 2026.
- You embed a third-party generative model in your own product and sell it under your name. You may count as the provider of the resulting system, so ask your model supplier what marking it already does.
- You publish AI-generated images, audio or video in the EU. Your disclosure duty as a deployer already applies.
- You offer image or video generation or editing tools. Check what safeguards you have against intimate-content and child-abuse outputs, because the new bans apply from 2 December.
Penalties
Penalties are set in Article 99. Breaking the bans in Article 5 can lead to fines of up to EUR 35 million or 7% of worldwide annual turnover, whichever is higher. Breaking the transparency duties in Article 50 falls in the next tier, up to EUR 15 million or 3%. For small and medium-sized businesses and small mid-caps, the lower of the two amounts applies.
A short plan for the next few weeks
- Work out whether you are a provider, a deployer, or both, for each AI feature you offer.
- For each generative feature, write down the date it was first placed on the EU market.
- Ask your model or tool suppliers in writing how they mark output and whether they plan to meet 2 December 2026.
- Check your own published content for AI-generated images, audio and video that need a disclosure.
- If you offer image or video generation, document the safeguards you have against the banned uses.
What this guide does not cover
The AI Act has more rules than these. If you are not sure whether the Act applies to you at all, start with the EU AI Act explained or which AI laws apply to you. For the high-risk rules that now start in December 2027, see Annex III: the high-risk AI use cases.
The binding text is on EUR-Lex: the AI Act and the Omnibus Regulation 2026/1744. Check it, or ask a qualified lawyer, before you rely on any date in this guide.
