Emotion‑recognition outlawed
- The EU AI Act now bans emotion‑recognition systems in the workplace, making those analytics illegal under Article 5(1)(f). (uctoday.com) - Companies that violate the ban face fines up to €35 million or 7% of global turnover. (uctoday.com) - Vendors and employers that sell or deploy such workplace analytics now face clear regulatory risk and will need to alter products or use. (uctoday.com)
The European Union now bans artificial intelligence systems that infer workers’ emotions from faces, voices, or other signals at work. (eur-lex.europa.eu) The ban sits in Article 5 of the Artificial Intelligence Act, Regulation (EU) 2024/1689, which was published in the Official Journal on July 12, 2024. The European Commission’s AI Act page lists “emotion recognition in workplaces and education institutions” among the law’s prohibited practices. (eur-lex.europa.eu) (digital-strategy.ec.europa.eu) The Commission published detailed guidelines on prohibited practices on February 4, 2025. Those guidelines say they are meant to support consistent enforcement across the bloc, while final legal interpretation belongs to the Court of Justice of the European Union. (digital-strategy.ec.europa.eu) Emotion-recognition software tries to guess feelings such as stress, boredom, or engagement from biometric or behavioral cues. The AI Act treats that use in offices, call centers, warehouses, and classrooms as an unacceptable-risk practice rather than a merely high-risk one. (digital-strategy.ec.europa.eu) (eur-lex.europa.eu) That classification changes the compliance question for employers and software vendors. A product sold as a tool for manager coaching, agent scoring, or productivity monitoring can cross into a banned category if it claims to read workers’ emotional state on the job. (digital-strategy.ec.europa.eu 1) (digital-strategy.ec.europa.eu 2) The AI Act uses a staged timeline, not a single start date for every rule. The Commission says the prohibitions on unacceptable-risk practices are already applicable, while many other AI Act obligations arrive later, on August 2, 2026 or August 2, 2027. (eur-lex.europa.eu) (digital-strategy.ec.europa.eu) The penalty ceiling is among the toughest in the law. The Commission’s AI Act questions page says infringements of prohibited practices can draw administrative fines of up to €35 million or 7% of worldwide annual turnover, whichever is higher. (digital-strategy.ec.europa.eu) The law does not ban every use of AI in hiring or management. The EU still treats some employment tools as high-risk systems, including tools used for recruiting, promotion, termination, task allocation, and performance monitoring, which means they can be used only under stricter obligations rather than being banned outright. (eur-lex.europa.eu) For vendors still marketing “sentiment,” “engagement,” or “mood” analytics to employers in Europe, the legal problem is now basic: if the software is reading emotions in the workplace, the European Union has moved that practice out of compliance checklists and into the banned column. (digital-strategy.ec.europa.eu)
