The morning read on Europe's digital policy Friday 16 October 2026
Search

AI Act prohibited practices: the full Article 5 list, including nudify apps

Article 5 of the EU AI Act bans eight categories of AI practice outright: harmful manipulation, exploitation of vulnerabilities, social scoring, crime prediction based only on profiling, untargeted scraping of facial images, emotion recognition at work and in schools, biometric categorisation by sensitive traits, and real-time remote biometric identification in public spaces for law enforcement (with narrow exceptions). As of October 2026, a ninth and tenth ban have been written into the law: AI systems that generate non-consensual intimate imagery ("nudify" tools) and child sexual abuse material, which apply from 2 December 2026. Breaching any of these bans can cost up to EUR 35 million or 7% of worldwide annual turnover, whichever is higher.

Where the bans come from and when they apply

The prohibitions sit in Article 5 of Regulation (EU) 2024/1689, the AI Act. They were the first part of the Act to bite:

  • 2 February 2025: the original eight prohibitions started to apply.
  • 2 August 2025: the penalty regime for breaching them became applicable.
  • 2 December 2026: the new bans on non-consensual intimate imagery and child sexual abuse material apply, added by the Digital Omnibus on AI, Regulation (EU) 2026/1744, which was published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026.

Each ban covers three acts: placing the system on the EU market, putting it into service, and using it. That means both providers (developers and vendors) and deployers (organisations using the system) can be liable. For the full sequence of AI Act deadlines, see our EU AI Act timeline.

The full list of prohibited AI practices

1. Subliminal, manipulative or deceptive techniques (Art. 5(1)(a))

Banned: AI that uses techniques beyond a person's awareness, or purposefully manipulative or deceptive techniques, to materially distort behaviour by impairing the ability to make an informed decision, in a way that causes or is reasonably likely to cause significant harm. Ordinary persuasion and lawful advertising are not caught; the test is material distortion plus significant harm.

2. Exploiting vulnerabilities (Art. 5(1)(b))

Banned: AI that exploits vulnerabilities linked to a person's age, disability, or specific social or economic situation to materially distort their behaviour in a way that causes or is likely to cause significant harm. Systems aimed at children, or at people in a precarious financial situation, are the kind of case this ban is designed to catch.

3. Social scoring (Art. 5(1)(c))

Banned: evaluating or classifying people over time based on their social behaviour or personal characteristics, where the resulting score leads to detrimental treatment in unrelated contexts, or treatment that is unjustified or disproportionate to the behaviour. The ban applies to public and private actors alike.

4. Predicting crime from profiling alone (Art. 5(1)(d))

Banned: assessing the risk that a person will commit a criminal offence based solely on profiling or on personality traits and characteristics. Systems that support a human assessment already grounded in objective, verifiable facts directly linked to criminal activity are outside the ban.

5. Untargeted scraping of facial images (Art. 5(1)(e))

Banned: creating or expanding facial recognition databases through untargeted scraping of facial images from the internet or CCTV footage.

6. Emotion recognition at work and in education (Art. 5(1)(f))

Banned: inferring the emotions of people in the workplace or in educational institutions. There is an exception for systems intended for medical or safety reasons. Emotion recognition outside these two settings is not prohibited, but is treated as high-risk.

7. Biometric categorisation by sensitive traits (Art. 5(1)(g))

Banned: categorising individuals based on biometric data to deduce or infer their race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation. Labelling or filtering of lawfully acquired biometric datasets, including in law enforcement, is excluded.

8. Real-time remote biometric identification for law enforcement (Art. 5(1)(h))

Banned in principle: live facial recognition and similar real-time remote biometric identification in publicly accessible spaces for law enforcement purposes. It is allowed only where strictly necessary for one of three objectives:

  • a targeted search for specific victims of abduction, trafficking or sexual exploitation, and for missing persons;
  • preventing a specific, substantial and imminent threat to life or physical safety, or a genuine threat of a terrorist attack;
  • locating or identifying a suspect of serious crimes listed in Annex II of the Act, punishable by a custodial sentence of at least four years.

Each use requires prior authorisation by a judicial or independent administrative authority (urgent cases can be authorised afterwards, within 24 hours), and member states must adopt national law before police can rely on these exceptions at all.

9. Non-consensual intimate imagery, or "nudify" tools (new Art. 5(1)(ba))

From 2 December 2026, the Act bans AI systems that generate or manipulate realistic images, video or audio depicting an identifiable person's intimate parts, or that person engaged in sexually explicit activity, without that person's explicit consent. The consent standard is strict: it must be freely given, specific, informed, unambiguous and explicit.

10. Child sexual abuse material (new Art. 5(1)(bb))

From the same date, AI systems that generate or manipulate child sexual abuse material, as defined in Directive 2011/93/EU, are prohibited. The ban does not apply where a "without right" defence available under national law applies. For the separate debate on detecting such material in private messages, see our explainer on the EU CSAM regulation.

How the new nudification ban works in practice

The Omnibus adds two clarifying paragraphs that define who is caught by the new bans:

  • Providers are caught when generating such content is the system's intended purpose, or when that output is reasonably foreseeable and reproducible without significant modification and the provider has not put reasonable safeguards in place. General-purpose image generators therefore need working safeguards, not just terms of use.
  • Deployers are caught when they use a system for the purpose of generating or manipulating such material.
  • Minor edits are excluded: editing that does not increase the exposure of depicted intimate parts, or change the nature of depicted sexual activity, does not count as manipulation under the intimate-imagery ban.

The recitals of the Omnibus also place cartoonish or physically impossible depictions of a person's body, and material that does not depict identifiable people, outside the intimate-imagery ban. They add that member states may also treat this conduct as a criminal offence, in which case the same act should not be punished twice.

What is not on the list

Several practices often assumed to be banned are in fact regulated, not prohibited:

  • Post (non-real-time) remote biometric identification is a high-risk use, subject to safeguards.
  • Emotion recognition outside workplaces and schools is high-risk, not banned.
  • Deepfakes in general are covered by the transparency duties in Article 50 (labelling and disclosure), not by Article 5, unless they fall within the new intimate-imagery or child abuse bans.
  • Biometric verification (confirming that a person is who they claim to be, such as unlocking a phone) is not remote identification and is not banned.

Penalties and enforcement

Under Article 99(3), infringing Article 5 is the most heavily fined breach in the Act: up to EUR 35 million or 7% of total worldwide annual turnover for the preceding financial year, whichever is higher. For SMEs and start-ups, the lower of the two amounts applies, and the Omnibus extends that rule to small mid-cap companies. National market surveillance authorities enforce the bans, and member states set the detailed penalty rules. Because the two new bans are inserted into Article 5 itself, the same ceiling applies to them.

Official guidance

On 4 February 2025 the European Commission published guidelines on prohibited AI practices, with legal explanations and practical examples for each of the original eight bans. They are not binding; the final word on interpretation belongs to the Court of Justice of the EU. The guidelines predate the Omnibus and do not yet cover the intimate-imagery and child abuse bans. Organisations auditing their AI use for Article 5 risks will often run that exercise alongside the staff training duty described in our guide to AI literacy under Article 4. More coverage is in our AI Act section.

Frequently asked questions

What AI practices are prohibited under Article 5 of the AI Act?

Article 5 bans harmful manipulation, exploitation of vulnerabilities, social scoring, crime prediction based solely on profiling, untargeted facial image scraping, emotion recognition at work and in schools, biometric categorisation by sensitive traits, and real-time remote biometric identification for law enforcement except in narrow cases. From 2 December 2026 it also bans AI systems that generate non-consensual intimate imagery or child sexual abuse material.

When did the AI Act bans take effect?

The original eight prohibitions have applied since 2 February 2025, and fines for breaching them since 2 August 2025. The two bans added by Regulation (EU) 2026/1744 apply from 2 December 2026.

Are nudify apps banned in the EU?

Yes, from 2 December 2026 the AI Act prohibits placing on the market or using AI systems that create realistic sexual images, video or audio of an identifiable person without their explicit consent. Providers of general tools are caught if such output is foreseeable and they lack reasonable safeguards.

What is the fine for using a prohibited AI system?

Up to EUR 35 million or 7% of total worldwide annual turnover, whichever is higher. For SMEs, start-ups and, after the Omnibus, small mid-cap companies, the lower of the two figures is the cap.

Is facial recognition banned under the AI Act?

Not entirely. Real-time remote biometric identification in public spaces for law enforcement is banned except for three listed objectives with prior authorisation, and building face databases through untargeted scraping is banned; other uses such as verification or retrospective identification are regulated rather than prohibited.