Prohibited AI Practices Under Article 5 Explained

The EU AI Act is risk-based: the higher the risk to health, safety and fundamental rights, the heavier the obligation. The four tiers are prohibited, high-risk, limited-risk (transparency) and minimal. Article 5 defines the top tier — practices judged to carry an unacceptable risk that no safeguards can adequately contain.

These practices are not regulated; they are forbidden. There is no compliance route that makes them lawful, and the ban applies extraterritorially — if the system or its output is used in the EU, it is caught wherever the provider sits.

Article 5 prohibits the placing on the market, putting into service or use of AI systems that involve:

• Harmful manipulation and deception — subliminal, manipulative or deceptive techniques that materially distort behaviour and cause significant harm.
• Exploitation of vulnerabilities — targeting people's age, disability or specific social or economic situation to distort their behaviour.
• Social scoring — evaluating or classifying people over time based on behaviour or personal traits, leading to detrimental or disproportionate treatment.
• Predictive policing of individuals — assessing the risk of someone committing a crime based solely on profiling or personality traits.
• Untargeted facial scraping — creating or expanding facial-recognition databases through untargeted scraping of images from the internet or CCTV.
• Emotion recognition in the workplace and education, except for medical or safety reasons.
• Biometric categorisation that infers sensitive traits such as race, political opinions, trade-union membership, religion or sexual orientation.
• Real-time remote biometric identification in publicly accessible spaces for law-enforcement purposes, subject to narrow, authorised exceptions.

Several of these carry limited, tightly drawn exceptions — but the default is a hard prohibition.

Under the phased timeline in Article 113, the Article 5 prohibitions were the first provisions to apply — from 2 February 2025, alongside the AI literacy obligations.

That matters for two reasons:

• The other deadlines are later, but this one has passed. GPAI obligations applied from 2 August 2025; Article 50 transparency applies from 2 August 2026; high-risk (Annex III) from 2 December 2027. The prohibitions are already live. See the full deadlines timeline.
• There is no grandfathering for banned practices. Any prohibited system on the market must be withdrawn now, not phased out later.

Article 5 breaches attract the Act's highest penalty band under Article 99:

• Up to €35 million, or
• Up to 7% of total worldwide annual turnover for the preceding financial year,

whichever is higher. This exceeds the top tier of GDPR fines and dwarfs the penalties for other AI Act obligations, which cap at €15M or 3% for most breaches and €7.5M or 1% for supplying incorrect information. For a full breakdown see the penalties guide.

Screening for prohibited practices should be the first step in any AI compliance review — because a system that is banned cannot be fixed:

• Inventory every AI system in use and in development, including third-party and embedded tools.
• Flag anything touching biometrics, emotion inference, behavioural scoring or manipulation for detailed Article 5 review.
• Scrutinise HR and edtech tools specifically — emotion recognition in workplaces and education is a common inadvertent breach.
• Check vendor claims: 'AI-powered engagement' or 'sentiment' features can quietly cross the line.
• Document your screening so you can demonstrate diligence to authorities.

Run your systems through the free risk checker to flag Article 5 exposure quickly, then read the complete guide for the wider programme.