The European Union’s Artificial Intelligence (AI) Act has come into effect, prohibiting certain hazardous AI applications effective February 2025. The new regulations aim to safeguard people’s safety and fundamental rights from the negative impacts of flawed or biased AI systems.
Flawed face recognition systems have led to incorrect identifications, including imprisoning individuals mistakenly labeled as offenders. A Dutch child care benefit scandal highlighted the risks of biased algorithms causing financial ruin for innocent people. To prevent such detrimental consequences, the EU has set red lines for AI practices in its rulebook.
Prohibited are AI systems manipulating people’s behavior or exploiting vulnerabilities, as well as certain social scoring methods unrelated to data context. Face recognition databases and live face recognition by police in public spaces are also partially banned. However, predictive policing based on objective facts and emotion recognition systems in education are allowed under specific conditions.
A loophole exists due to national security exemptions, which may lead to AI systems being exported to third countries despite EU prohibitions. Those who believe their rights have been infringed can file complaints with market surveillance authorities, but the effectiveness of enforcement remains uncertain.
Source: https://algorithmwatch.org/en/ai-act-prohibitions-february-2025