CLUSTER / THE DECISION MAP
Prohibited practices (Article 5).
The practices Article 5 prohibits, the conditions and exceptions in outline, and how the risk-based approach works.
IN THIS CLUSTER
12 guides.
Prohibited practices under Article 5: the practices the EU AI Act bans outright
The Article 5 prohibitions in outline, their application dates, the narrow exceptions, and the screening order that keeps a prohibited practice from being treated as a risk tier.
Read guide ↗02 / CLASSIFYUnderstand AI Act risk categories
How to approach prohibited practices, high-risk AI, transparency obligations and lower-risk uses without assuming a label from a single feature.
Read guide ↗CLUSTER ARTICLEHow the AI Act risk-based approach works: prohibited, high-risk, transparency and other uses
The risk-based structure of the Act explained as a screening order, so that each system is tested against the right tier in the right sequence.
Read guide ↗CLUSTER ARTICLEManipulative techniques: Article 5(1)(a) and subliminal, manipulative or deceptive AI
The first Article 5 prohibition explained element by element, including material distortion of behaviour, significant harm, and the line between persuasion and prohibition.
Read guide ↗CLUSTER ARTICLEExploitation of vulnerability: Article 5(1)(b) and AI that targets age, disability or circumstance
How Article 5(1)(b) treats AI systems that exploit vulnerabilities, what counts as a vulnerability, and how intention and harm are tested.
Read guide ↗CLUSTER ARTICLESocial scoring: Article 5(1)(c) and AI scores that punish people in unrelated contexts
How Article 5(1)(c) treats social scoring by public and private actors, the two alternative conditions, and the difference from ordinary credit or eligibility assessment.
Read guide ↗CLUSTER ARTICLEPredictive policing: Article 5(1)(d) and crime risk assessment based solely on profiling
The narrow prohibition on assessing criminal risk solely from profiling or personality traits, and the objective-facts exception that must be read with care.
Read guide ↗CLUSTER ARTICLEFacial image scraping: Article 5(1)(e) and untargeted collection for facial recognition databases
The prohibition on creating or expanding facial recognition databases by untargeted scraping of facial images from the internet or CCTV, and what targeted collection means.
Read guide ↗CLUSTER ARTICLEEmotion inference at work and in education: Article 5(1)(f) and its medical or safety exception
The prohibition on inferring emotions of workers and learners, what counts as emotion inference, and how the narrow medical or safety exception is tested.
Read guide ↗CLUSTER ARTICLEBiometric categorisation: Article 5(1)(g) and inferring sensitive attributes from biometric data
How Article 5(1)(g) prohibits biometric categorisation that infers race, political opinion, trade union membership, religion, sex life or sexual orientation, and its narrow carve-outs.
Read guide ↗CLUSTER ARTICLEReal-time remote biometric identification: Article 5(1)(h) and the narrow law enforcement exceptions
The prohibition on real-time remote biometric identification in publicly accessible spaces for law enforcement, its three purposes, and its authorisation safeguards.
Read guide ↗CLUSTER ARTICLENew prohibitions from 2 December 2026: Article 5(1)(ba) and (bb)
The two prohibitions added by Regulation (EU) 2026/1744, their provider and deployer conditions, the safeguards test, and what changes on 2 December 2026.
Read guide ↗