EU AI Act Article 5: Prohibited AI Practices

Article 5 of Regulation (EU) 2024/1689 lists prohibited AI practices. The original points (a) through (h), applicable since February 2, 2025, address specified harmful manipulation or deception; exploitation of defined vulnerabilities; social scoring that leads to harmful treatment; crime-risk assessments based solely on profiling or personality traits; untargeted facial-image scraping for recognition databases; emotion inference at work or in education except for medical or safety purposes; biometric categorisation to infer listed sensitive traits; and most real-time remote biometric identification by law enforcement in public spaces. Each rule has its own elements and exceptions. Article 5 does not ban every tool used in these areas. Its specific thresholds matter: points (a), (d), and (h), for example, define harm conditions, an evidential exception, and authorization safeguards. Other EU or national laws may still apply. Regulation (EU) 2026/1744, published in July 2026 and in force, inserted points (ba) and (bb). They address systems generating or manipulating realistic intimate or sexually explicit material of an identifiable person without explicit consent, and certain child sexual abuse material. The amendment also specifies provider and deployer conditions and safeguards; it does not make every general-purpose image or audio model prohibited. These two points and related paragraphs 1a and 1b apply from December 2, 2026, so they are enacted but not yet applicable on September 26, 2026. For a real assessment, map the system, actor, intended purpose, output, affected person, relevant Article 5 point, thresholds, and exceptions. Article 5 is not the Act’s high-risk classification test, and it is not an exhaustive list of every unlawful AI use. Check the current consolidated text and other applicable law. This guide is informational, not legal advice.