Starting today, February 2, 2025, the European Union will begin enforcing its first set of prohibitions under the AI Act, targeting AI systems deemed an “unacceptable risk.” The law, which took effect last August, is the world’s first comprehensive framework regulating AI development and use. Companies violating these bans face fines of up to €35 million or 7% of global revenue.
Key Takeaways:
- AI systems that manipulate users, exploit vulnerabilities, or use real-time biometric surveillance in public spaces are now banned.
- Violators could face severe penalties, even if their headquarters are outside the EU.
- Some exceptions exist for law enforcement and medical applications with strict oversight.
- Additional guidelines are expected later in 2025 to clarify enforcement and compliance.
The AI Act classifies AI systems into four risk categories: minimal, limited, high, and unacceptable. The latest compliance deadline applies to AI practices that fall into the highest risk category—those that threaten human rights, privacy, or public safety. Among the banned applications are predictive policing based on profiling, emotion recognition in workplaces and schools, and AI systems that scrape biometric data from public sources.
The European Commission will oversee enforcement in partnership with national regulators. While fines won’t be imposed immediately, companies operating in the EU must demonstrate compliance or risk penalties later this year. Some industries, including financial services, healthcare, and social media, are under particular scrutiny due to potential risks related to AI-driven decision-making.
The EU plans to implement further regulatory measures, including obligations for high-risk AI systems in 2026. Meanwhile, global tech firms will need to navigate this shifting regulatory landscape carefully to maintain compliance and continue operating in Europe.