Article 5 of the EU AI Act prohibits AI practices that materially distort a person’s behavior and cause, or are reasonably likely to cause, significant harm. It covers two routes: manipulative techniques and exploitation of vulnerabilities. Manipulation may be subliminal or openly deceptive; visibility does not mean a person can resist it. The technique-based prohibition also requires appreciable impairment of the person’s ability to make an informed decision. section 7.2.2
Assess the practice by asking:
- Does it use subliminal, purposefully manipulative, or deceptive techniques that distort behavior and meet the harm threshold? section 7.2.2
- Does it exploit vulnerability due to age, disability, or a specific social or economic situation, with the required distortion and harm? section 7.2.2
Targeting under pressure
An insurance system flags elderly policyholders showing signs of cognitive decline and prompts agents to use urgency-based scripts to sell a complex product. A chatbot that invokes a user’s children to manufacture urgency for a high-interest loan illustrates visible but hard-to-recognize manipulation.
