Prohibited AI Practices Under the EU AI Act: What’s Already Banned Since 2025
Last updated: 2 November 2026
Key takeaways
- Since 2 February 2025, eight uses of AI have been banned in the EU, and since 2 August 2025 breaches can cost up to EUR 35 million or 7% of worldwide annual turnover.
- The bans apply to providers and deployers, including those outside the EU whose AI output is used in the EU. A deployer cannot rely on its vendor’s contract terms.
- Several bans catch ordinary commercial products: emotion recognition in HR and education tools, manipulative marketing that causes significant harm, and AI that rates people as likely criminals based only on their profile.
- Fact-based transaction monitoring is outside the ban on predicting crime from profiles.
- From 2 December 2026, the Digital Omnibus adds two new bans on AI-generated non-consensual intimate images and child sexual abuse material.
Since 2 February 2025, eight uses of AI have been banned in the EU outright, and since 2 August 2025 breaking these bans can cost up to EUR 35 million or 7% of worldwide annual turnover. Many businesses assume the list is about police surveillance and dystopian scenarios that have nothing to do with them. In fact, several of the bans catch ordinary commercial products: HR tools, marketing chatbots, exam software and image generators.
Read more: EU AI Act Explained: The Four Risk Tiers
Read more: EU AI Act Explained: The Four Risk Tiers
Who the bans apply to
The bans apply to providers that place AI systems on the EU market, including providers based outside the EU, and to deployers in the EU, as well as to providers and deployers outside the EU whose AI output is used in the EU. The AI Act does not apply to systems used exclusively for military, defence or national security purposes or to AI developed and put into service solely for scientific research, and it imposes no obligations on individuals using AI for purely personal, non-professional purposes. A deployer cannot hide behind its vendor’s terms: according to the Commission’s Guidelines, an employer that uses AI to infer emotions at the workplace is caught by the ban even if the provider excluded that use in its contract.
Manipulation and exploitation of vulnerabilities
The first two bans cover AI that uses subliminal, purposefully manipulative or deceptive techniques, and AI that exploits vulnerabilities due to age, disability or a specific social or economic situation. In both cases the AI must materially distort a person’s behaviour in a way that causes, or is likely to cause, significant harm. The Commission’s examples include a chatbot that impersonates a person’s friend or relative with a synthetic voice, AI toys that exploit the vulnerability of children, and AI that targets older people with deceptive personalised offers or scams. Ordinary persuasion and personalised advertising are not banned. The test is significant harm, which is why marketing teams should document how they avoid it.
Social scoring
AI may not evaluate or classify people based on their social behaviour or personal traits where the resulting score leads to detrimental treatment in unrelated contexts, or treatment that is unjustified or disproportionate. The ban applies to private companies as well as public authorities. It does not affect lawful evaluation carried out for a specific purpose in line with EU and national law. Credit scoring, for example, is generally treated as high-risk rather than banned, unless it meets the conditions of the social scoring ban.
Predicting crime from profiles
AI may not assess the risk that a person will commit a crime based solely on profiling or on their personality traits. The exception is AI that supports a human assessment already based on objective and verifiable facts directly linked to criminal activity. This matters for compliance and anti-money laundering teams. Transaction monitoring that flags suspicious transactions based on facts is outside the ban. A tool that rates customers as likely criminals based only on their profile is inside it.
Facial databases, emotions at work and sensitive traits
Three more bans concern biometrics. AI may not build facial recognition databases through untargeted scraping of images from the internet or CCTV footage. It may not infer emotions in the workplace or in education, except for medical or safety reasons. And it may not use biometric data to categorise people by race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation, although labelling or filtering lawfully acquired biometric datasets, including in law enforcement, is allowed.
The emotion ban is the one most likely to catch ordinary products: call centre software that scores agents’ emotions, or proctoring tools that read students’ faces during exams. Note its limits, though. Emotion recognition of customers is not banned, but it is a high-risk use, and since 2 August 2026 people exposed to it must be informed.
Real-time facial recognition by the police
The last ban covers real-time remote biometric identification in publicly accessible spaces for law enforcement. It has three narrow exceptions: the targeted search for victims of abduction, trafficking or sexual exploitation and for missing persons; the prevention of a specific, substantial and imminent threat to life or physical safety, or of a genuine and present or foreseeable threat of a terrorist attack; and the search for suspects of serious crimes punishable by at least four years in prison. Each use needs prior authorisation by a judicial or independent administrative authority; in urgent cases, use may start without it, but authorisation must be requested within 24 hours.
Two new bans from 2 December 2026
The Digital Omnibus adds two prohibitions: AI that generates or manipulates realistic intimate images or videos of an identifiable person without their explicit consent, and AI that generates child sexual abuse material. Providers are caught where this is the system’s intended purpose, or where such output is a reasonably foreseeable and reproducible result and adequate safeguards are missing. Deployers are caught when they use the system to generate such material themselves. For providers of image and video generators, the practical question is whether their safeguards will be in place by December.
Fines and enforcement
Breaches can lead to fines of up to EUR 35 million or 7% of worldwide annual turnover, whichever is higher. For SMEs and start-ups, the lower of the two amounts applies. Enforcement lies with national market surveillance authorities, which Member States had to designate by 2 August 2025. Several countries were late: Germany, for example, made the Bundesnetzagentur its central AI market surveillance authority only in July 2026. As of October 2026, we have not identified any publicly reported fine under the bans, but the authorities are now in place.
The Commission’s Guidelines
The Commission’s Guidelines on prohibited AI practices had their content approved on 4 February 2025, and the final Communication is dated 29 July 2025. They are not binding, and only the Court of Justice of the EU can interpret the AI Act authoritatively. Still, they are the best indication of how authorities will read the bans, and their examples are a practical checklist.
What we recommend
- Screen every AI system you build or use against the list, including features inside larger products.
- Check HR, workplace and education tools for any function that infers emotions.
- Review personalised marketing aimed at children, older people or financially vulnerable customers.
- If you offer image or video generation, put safeguards in place before 2 December 2026.
- Make sure your contracts with AI vendors do not leave you using a banned feature by default.
The bottom line
The banned practices are the shortest part of the AI Act and the most expensive one to get wrong. For most businesses, the risk is not a deliberate decision to use a banned system but a feature nobody checked. If you are not sure whether any of your AI tools come close to the line, AMLzone can help you review them and document your position.
Sources
1. Regulation (EU) 2024/1689 (AI Act), Art. 5, Art. 99(3) and Art. 113. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
2. AI Act, Art. 2(1), (3), (6) and (10). https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
3. European Commission, Guidelines on prohibited artificial intelligence practices established by Regulation (EU) 2024/1689, C(2025) 5052, para 14. https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-prohibited-artificial-intelligence-ai-practices-defined-ai-act
4. AI Act, Art. 5(1)(a)–(b); Commission Guidelines on prohibited AI practices, section on harmful manipulation and exploitation of vulnerabilities. https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-prohibited-artificial-intelligence-ai-practices-defined-ai-act
5. AI Act, Art. 5(1)(c) and recital 31; Commission Guidelines on prohibited AI practices, section on social scoring. https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
6. AI Act, Art. 5(1)(d). https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
7. AI Act, Art. 5(1)(e)–(g). https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
8. AI Act, Annex III, point 1(c), and Art. 50(3). https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
9. AI Act, Art. 5(1)(h), (2) and (3). https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
10. Regulation (EU) 2026/1744 (Digital Omnibus on AI), new Art. 5(1)(ba) and (bb), Art. 5(1a) and amended Art. 113. https://eur-lex.europa.eu/eli/reg/2026/1744/oj/eng
11. AI Act, Art. 99(3) and (6). https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng
12. European Commission, “Market surveillance authorities under the AI Act”; Bundesnetzagentur, press release of 29 July 2026. https://digital-strategy.ec.europa.eu/en/policies/market-surveillance-authorities-under-ai-act
13. European Commission, “Commission publishes the Guidelines on prohibited artificial intelligence (AI) practices, as defined by the AI Act”. https://digital-strategy.ec.europa.eu/en/library/commission-publishes-guidelines-prohibited-artificial-intelligence-ai-practices-defined-ai-act
02.11.2026
