EU AI Act — Article 5

In force since 2025-02-02

What it requires

The highest-stakes provision in AI manipulation law anywhere: Art.5(1)(a)–(b) bans subliminal, purposefully manipulative or deceptive techniques that materially distort behaviour and cause significant harm, plus the exploitation of vulnerabilities from age, disability or socio-economic situation. Art.5(1)(f) separately and explicitly prohibits emotion recognition in the workplace and in educational institutions — not merely flags it as high-risk. This is distinct from the Act's Annex III high-risk tier, which is domain-based and triggers conformity-assessment duties regardless of whether manipulation occurs.

Teeth

Up to €35 million or 7% of worldwide annual turnover (Art. 99(3)) — the Act's highest penalty tier.

Red lines it maps to

• RL1 — Anthropomorphic Intimacy Escalation
• RL3 — Sexual or Harmful Content to Minors
• RL4 — Emotion Recognition in Workplace or Education
• RL10 — Encouraging or Facilitating Violence or Crime

Primary source →