New prohibitions from 2 December 2026: Article 5(1)(ba) and (bb)

What does point (ba) cover, in general terms?

Point (ba) covers AI systems that generate or manipulate realistic images, videos, audio or similar material depicting an identifiable person's intimate parts, or depicting an identifiable person engaged in sexually explicit activities, without qualifying consent. The material must be realistic, the person must be identifiable, and the consent must be freely-given, specific, informed, unambiguous and explicit, for that generation or manipulation. A general consent to use a likeness, or consent to one use, does not automatically cover a different generation or manipulation. This guide describes the subject matter only in general terms, as the verified date list requires. No examples involving identifiable people are given, and no graphic detail is reproduced. The full wording in the consolidated text should be read before any screening decision.

What does point (bb) cover, in general terms?

Point (bb) covers AI systems that generate or manipulate material or performance within the meaning of Article 2, points (c) and (e), of Directive 2011/93/EU, which concern child sexual abuse material and pornographic performance as defined in that Directive, except where a without right defence applies under national law. The reference to the Directive means the scope is defined by that instrument, not by a general description in the AI Act alone. The exception for activities carried out under domestic legal powers, such as legitimate generation for criminal proceedings or for preventing, detecting or investigating crime, and legitimate red-teaming and evaluation aimed at assessing compliance with the prohibition itself, is described in the recitals to Regulation (EU) 2026/1744. This guide gives no further detail on the material, and any work in this area should be handled only by the authorities and specialists the law envisages.

For the broader regulatory framework, see prohibited practices under Article 5.

How do paragraphs (1a) and (1b) test providers and deployers?

Paragraph (1a) provides that placing on the market or putting into service is only prohibited for the new points where the generation or manipulation is the intended purpose of the system, or where the system's design, training, architecture, capabilities or user-facing functionalities make that generation or manipulation a reasonably foreseeable and reproducible outcome, without requiring significant technical modification, and the system does not have reasonable and adequate technical safety measures and other safeguards to reliably prevent it, taking into account reasonably foreseeable misuse, and to correct observed or reported misuse. Paragraph (1b) sets a corresponding test for use by deployers. The safeguards test is therefore central for general-purpose generative systems that were not built for the prohibited purpose but could foreseeably be misused to produce the covered material.

What safeguards are relevant to the test?

The recitals and the safeguards test point to measures such as data cleaning, refusal training, safe prompt design, output controls, runtime guardrails, content classification, abuse detection, and notice-and-action mechanisms for reported misuse. No single measure is sufficient by itself, and the test is whether the measures together are reasonable and adequate to reliably prevent the generation or manipulation and to correct observed or reported misuse, taking into account reasonably foreseeable misuse. A provider should record the measures in place, the testing performed, the misuse observed, and the corrections made, and should treat the safeguards as a living control that is reviewed when new misuse patterns appear. A deployer should record its own use controls and its response to reports of misuse.

What changes on 2 December 2026, and what does not?

On 2 December 2026, the two new prohibitions begin to apply under Article 113 as amended by Regulation (EU) 2026/1744, and the transitional deadline for certain providers of systems generating synthetic content already placed on the market before 2 August 2026 to comply with the Article 50(2) marking and detection obligation also falls on that date. The original Article 5 prohibitions are unaffected and have applied since 2 February 2025. The new prohibitions carry no grandfathering for systems already on the market in the sense described in paragraph (1a): a system already supplied must still meet the safeguards test where the foreseeable-outcome limb is met. Teams should therefore complete screening and safeguards work before the date, rather than treating it as a future compliance project that starts in December.

Continue with the risk-based approach.

What is outside the scope of this guide?

This guide describes the new prohibitions in general terms only, as required for this subject matter. It does not assess any particular system, model, safeguard or use, does not reproduce the definitions in Directive 2011/93/EU, and does not advise on criminal law, reporting duties or victim support, which are governed by other law and by the authorities responsible for it. Any suspected production of the material covered by point (bb) should be handled through the reporting and safeguarding routes the applicable law provides, and not through product experimentation. Borderline questions about consent, safeguards or intended purpose should be confirmed with a competent authority or legal adviser, and this guide should not be used as a basis for testing the limits of the prohibition.

Frequently asked questions

Continue through this AI Act cluster

• prohibited practices under Article 5
• the risk-based approach
• EU AI Act timeline from 2024 to 2028
• transparency duties for AI
• EU AI Act timeline
• Understand AI Act risk categories