AI Act 2 December 2026: New Article 5 Bans and the Article 50(2) Deadline for Existing Generative AI

On 2 December 2026 the EU AI Act starts to apply two new prohibitions in Article 5 and ends the four-month transition that gave providers of generative AI systems placed on the market before 2 August 2026 time to comply with Article 50(2). Both rules were added by Regulation (EU) 2026/1744, in force since 27 July 2026.

The AI Act is Regulation (EU) 2024/1689 (OJ L, 2024/1689, 12.7.2024), amended by Regulation (EU) 2026/1744 of 8 July 2026. Under the EU AI Act, the date matters to providers and deployers of AI systems that generate images, audio, video or text.

Compiled by the AIActGuide.net Editorial Team from the Official Journal text; how we check every statement: methodology.

What Applies under the AI Act from 2 December 2026?

Two sets of rules apply from 2 December 2026: Article 5(1), first subparagraph, points (ba) and (bb), with Article 5(1a) and (1b) (Article 113, third paragraph, point (a)), and Article 50(2) for generative AI systems placed on the market before 2 August 2026 (Article 111(4)). Three obligations result:

The other Article 5 points, (a), (b) and (c) to (h), apply since 2 February 2025, and Article 50 applies since 2 August 2026. Each step of the EU AI Act timeline carries its own date and provision.

The Two Prohibitions Added by Regulation (EU) 2026/1744

Regulation (EU) 2026/1744 inserted point (ba), which bans AI systems that generate or manipulate realistic intimate or sexually explicit material of an identifiable person without that person's freely-given, specific, informed, unambiguous and explicit consent, and point (bb), which bans AI systems that generate or manipulate child sexual abuse material. Point (bb) covers material within Article 2, points (c) and (e), of Directive 2011/93/EU, except where a "without right" defence applies under national law.

Recital 12 of Regulation (EU) 2026/1744 limits realism to credible depictions of a face, voice or body and excludes cartoonish depictions, and its recital 11 refers to "nudification" applications.

Article 5(1) now has ten lettered points: points (a), (b) and (c) to (h) apply since 2 February 2025, points (ba) and (bb) from 2 December 2026. The Commission's AI Act policy page (last updated 19 October 2026) states that the act "prohibits nine practices", while the prohibited AI practices under Article 5 are ten lettered points in the act itself.

Provider and Deployer Tests for the New Prohibitions

Article 5(1a) splits the new prohibitions by role: a provider breaches them when generating such material is the intended purpose or a reasonably foreseeable and reproducible outcome without reasonable and adequate safeguards, and a deployer breaches them when it uses the system for that purpose. The provider test has three limbs:

Article 5(1b) applies to point (ba). Whether safeguards are reasonable and adequate is, in our interpretation, a design question for the provider.

The Four-Month Transition for Generative AI Systems Placed before 2 August 2026

Providers of AI systems, including general-purpose AI systems, that generate synthetic audio, image, video or text and were placed on the market before 2 August 2026 must comply with Article 50(2) by 2 December 2026 (Article 111(4)). Article 50 applies since 2 August 2026 (Article 113, second paragraph), and Article 111(4) is its only transition. Recital 38 of Regulation (EU) 2026/1744 calls it "a transitional period of four months".

Article 50(2) requires outputs to be marked in a machine-readable format and detectable as artificially generated or manipulated. Among the transparency obligations under Article 50, it is the only one with a transition, and its technical solutions must be effective, interoperable, robust and reliable as far as technically feasible.

Which Providers Must Comply by 2 December 2026, and Which Duties Never Had a Transition

Article 111(4) covers only the provider duty to mark outputs in a machine-readable format and make them detectable under Article 50(2); no other Article 50 duty received a transition. Four duties never had one:

• Article 50(1): informing natural persons that they interact with an AI system.
• Article 50(3): informing persons exposed to emotion recognition or biometric categorisation systems.
• Article 50(4): disclosing deep fakes and AI-generated text published to inform the public on matters of public interest.
• Article 50(5): giving the information clearly and distinguishably, at the latest at first interaction or exposure.

Footnote 50 of the Commission guidelines on Article 50 (content approved and published on 20 July 2026) states that the grandfathering of Article 111(2) does not apply to Article 50, and the open-source exclusion of Article 2(12) does not reach Article 50 or Article 5 systems. The watermarking duties of providers under Article 50(2) differ from the labelling duties of deployers under Article 50(4).

What the Transparency Code of 10 June 2026 Adds for Providers

The Code of Practice on Transparency of AI-generated Content, final since 10 June 2026, is the voluntary tool providers can sign to show how they meet Article 50(2); it also covers Article 50(4) and (5). The Commission concluded on 8 July 2026 that the code is adequate, and the AI Board assessed it on 9 July 2026.

For providers, the code sets at least two marking layers (digitally signed metadata and an imperceptible watermark), a watermark for free-form text longer than 200 tokens, detection free of charge with a fee possible only for signatories below 1,000,000 monthly users, and detection interoperability by 2 February 2027 (Measure 3.4(c)).

How Regulation (EU) 2026/1744 Set the 2 December 2026 Milestone

The 2 December 2026 milestone exists because Regulation (EU) 2026/1744, the Digital Omnibus on AI, inserted both the new prohibitions and Article 111(4) into the AI Act. Its Article 1, points (7), (39)(b) and (40), made the changes:

Points (ba) and (bb) were not in the Commission proposal COM(2025) 836 of 19 November 2025; the co-legislators added them. The Digital Omnibus on AI changed the proposal's date for Article 50(2) to 2 December 2026.

Before 27 July 2026, Article 113, third paragraph, point (a) had no exception for Article 5. The consolidated text 02024R1689-20260727 has no legal effect; the Official Journal texts are authentic.

Authorities and Maximum Amounts for Breaches from 2 December 2026

National market surveillance authorities enforce both obligations from 2 December 2026, except where the AI Office is exclusively competent under Article 75(1) as replaced. That competence covers AI systems based on a general-purpose AI model of the same provider or undertaking (except Annex I products and Annex III point 2, Article 74(6) and Annex III point 8 systems) and AI systems that constitute or are integrated into a very large online platform or search engine designated under Regulation (EU) 2022/2065.

For SMEs, including start-ups, each maximum is the lower of the two amounts (Article 99(6)); for small mid-caps this applies only under Article 99(4) and (5) (Article 99(6a)). The EU AI Act penalties span Articles 99 to 101.

What Providers and Deployers Should Do before 2 December 2026

Five steps prepare an organisation for 2 December 2026, starting with knowing which AI systems it provides or deploys and when each was placed on the market. The steps are general and do not assess any individual system:

1. List generative systems and the date each was placed on the EU market.
2. Implement machine-readable marking and detection, both elements.
3. Test whether generating such material is an intended purpose or a reasonably foreseeable and reproducible outcome, and document the safeguards.
4. Restrict deployer use of the system for that purpose.
5. Decide on signing the transparency code and plan for its detection interoperability measure.

A generative system placed on the market since 2 August 2026 is already bound by Article 50(2), because Article 111(4) covers only earlier systems.

The EU AI Act checklist lists every obligation by role.

After 2 December 2026: The Next AI Act Milestones

The next AI Act application date after 2 December 2026 is 2 August 2027, when general-purpose AI models placed on the market before 2 August 2025 must comply (Article 111(3)) and each Member State must have at least one national AI regulatory sandbox operational (Article 57(1) as amended).

Chapter III, Sections 1 to 3, apply to Annex III high-risk AI systems from 2 December 2027 (Article 113, third paragraph, point (c)(i)) and to Annex I systems from 2 August 2028 (point (c)(ii)). The overview of every application date of the EU AI Act runs from 2024 to 2030.

AI Act 2 December 2026: Related Questions

Article 2(1), Article 111 and Article 113 of the AI Act settle the common questions about 2 December 2026.

Is the AI Act Delayed?

Partly: Regulation (EU) 2026/1744 moved Chapter III, Sections 1 to 3, to 2 December 2027 for Annex III systems and 2 August 2028 for Annex I systems, while the 2 December 2026 obligations are new. The Digital Omnibus on AI and the new high-risk dates explain what it changed.

Has the AI Act Been Passed and Implemented?

Yes: Regulation (EU) 2024/1689 was published on 12 July 2024, entered into force on 1 August 2024 and applies in stages, of which 2 December 2026 is one. Regulation (EU) 2026/1744 of 8 July 2026 amended it with effect from 27 July 2026.

Do the 2 December 2026 Rules Apply to Providers outside the EU?

Yes: both obligations reach providers established outside the EU that place AI systems on the EU market, and providers and deployers outside the EU whose output is used in the EU (Article 2(1), points (a) and (c)). The guide to EU AI Act for non-EU companies sets out when an authorised representative is needed.

Do the New Bans Reach AI Systems Placed on the Market before 2 December 2026?

Yes: Article 111(1) and (2) apply without prejudice to Article 5, so the new prohibitions bind every AI system from 2 December 2026, whenever it was placed on the market. Of the 2 December 2026 obligations, only Article 50(2) had a transition, under Article 111(4).

Has the Commission Published Guidance on the New Bans?

No update of the Commission guidelines on prohibited AI practices (C(2025) 5052) covering points (ba) and (bb) is published on the Commission library page as of 24 October 2026. That library page was last updated on 31 July 2025. The EU AI Act guidelines are listed with the date of each document.