Prohibited AI practices
Article 5 bans certain AI uses outright, including social scoring, untargeted facial-image scraping, most emotion recognition at work or school, and certain manipulative or exploitative systems. The ban has applied since 2 February 2025.
Prohibited practices carry the heaviest penalties in the Act — up to €35 million or 7% of total worldwide annual turnover. The list covers subliminal or purposefully manipulative techniques that materially distort behaviour and cause significant harm, exploitation of vulnerabilities linked to age, disability or social or economic situation, social scoring leading to detrimental treatment in unrelated contexts, individual criminal-offence risk prediction based solely on profiling or personality traits, untargeted scraping of facial images to build recognition databases, emotion inference in the workplace and in education, biometric categorisation to infer sensitive attributes, and real-time remote biometric identification in public spaces for law enforcement outside narrow authorised cases.
- Art. 5
- Art. 99(3)
Prohibited AI practices: why it matters
This is the only tier where compliance is not available: there is no conformity assessment, no documentation package and no consent mechanism that makes a prohibited practice lawful. It is also the tier that has been enforceable the longest, since 2 February 2025, so exposure here is present-tense rather than a 2027 planning item. The prohibitions bind deployers as directly as providers, which means buying the capability from a vendor transfers none of the risk.
Prohibited AI practices in practice
A retail chain installs cameras that flag which shoppers look frustrated and routes staff accordingly — outside the workplace this is not automatically prohibited, but the moment the same system is pointed at the checkout staff to measure their mood during shifts, it is emotion inference in the workplace and falls under the ban. The hardware, vendor and model are unchanged; only the population being analysed moved, and that is what makes the practice illegal.
Prohibited AI practices: the mistake people make
Assuming employee consent or a legitimate business purpose can rescue a prohibited practice. Article 5 is a flat prohibition, not a balancing test — unlike the GDPR there is no lawful basis to point at, and the narrow exceptions that do exist (medical or safety grounds for emotion inference, judicially authorised law-enforcement biometrics) are drawn deliberately tightly.
Prohibited AI practices: common questions
- When did the EU AI Act prohibitions take effect?
- 2 February 2025. The Article 5 bans and the Article 4 AI-literacy duty were the first provisions to apply, six months after the Act entered into force, and well ahead of the high-risk regime.
- Is emotion recognition banned everywhere in the EU?
- No. Article 5 prohibits inferring emotions in the workplace and in education institutions, with narrow exceptions for medical or safety reasons. In other contexts, emotion recognition is generally treated as high-risk and carries transparency duties rather than being banned.
- What is the penalty for a prohibited AI practice?
- Administrative fines of up to €35 million or 7% of total worldwide annual turnover for the preceding financial year, whichever is higher. It is the highest penalty band in the regulation.
Prohibited AI practices: terms that come up alongside it
Run the free 2-minute EU AI Act checker — no sign-up.