Definition · Plain-language
Prohibited AI practices
Prohibited AI practices are the AI uses banned under Article 5 of the EU AI Act because they are judged to pose an unacceptable risk to safety, livelihoods and rights.
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
The unacceptable-risk tier
Within the EU AI Act’s risk-based structure, prohibited practices occupy the highest tier: unacceptable risk. These are uses considered a clear threat to people’s safety, livelihoods and fundamental rights, and so are banned rather than merely regulated. Set out in Article 5, the prohibitions apply across the Union from an early stage in the Act’s phased timeline, taking effect from February 2025. They represent a relatively small but significant list, drawing a line that no compliance measures can cross for the uses concerned.
Examples of banned uses
Article 5 prohibits, among others, AI that deploys subliminal or purposefully manipulative techniques or exploits vulnerabilities of specific groups in ways likely to cause significant harm; social scoring by public authorities that leads to unjustified or disproportionate detrimental treatment; the untargeted scraping of facial images from the internet or CCTV to build facial-recognition databases; certain biometric categorisation inferring sensitive attributes; emotion recognition in the workplace and educational settings, with limited exceptions; and certain predictive-policing uses based solely on profiling. Real-time remote biometric identification in publicly accessible spaces for law enforcement is banned subject to narrowly defined exceptions.
Scope and exceptions
The prohibitions are drafted with specific conditions and, in some cases, narrow exceptions, so the precise boundary of each ban depends on the detailed wording of Article 5. The most discussed exceptions concern real-time remote biometric identification by law-enforcement authorities, which is permitted only in tightly defined situations subject to safeguards and authorisation. Because these are the bans carrying the heaviest penalties under the Act — up to €35 million or 7% of global annual turnover — the exact scope of each prohibited practice is read closely against the legal text.
Key facts
At a glance
- Definition: AI uses banned under Article 5 of the EU AI Act.
- Risk tier: Unacceptable risk — the highest tier.
- Examples: Social scoring, manipulative systems, untargeted facial scraping.
- Biometrics: Narrow exceptions for real-time remote biometric ID by law enforcement.
- Applies from: February 2025.
- Max penalty: Up to €35M or 7% of global annual turnover.
Common misconceptions
What people often get wrong
Often heard: Prohibited AI practices cover most everyday business uses of AI.
Actually: Article 5 bans a specific, relatively short list of unacceptable-risk uses. Most ordinary AI applications fall into the high-risk, limited-risk or minimal-risk tiers rather than being prohibited.
Often heard: All forms of biometric identification are banned by the EU AI Act.
Actually: The Act prohibits certain biometric uses and restricts real-time remote biometric identification by law enforcement, but with narrowly defined exceptions. Not all biometric uses are prohibited.
Often heard: The prohibitions only take effect with the rest of the Act in 2026.
Actually: The Article 5 prohibitions are among the earliest provisions to apply, taking effect from February 2025, ahead of the high-risk regime that largely applies from August 2026.
Going deeper
