The European Union Artificial Intelligence Act, passed in 2024 and entering enforcement in phases through 2025 and 2026, is the world’s first comprehensive attempt to regulate AI through binding legislation. After years of debate, lobbying, and last-minute amendments, it is now law.
The Act takes a risk-based approach, imposing different requirements depending on how an AI system is used. At the highest level, it prohibits certain AI applications outright: social scoring systems, real-time biometric surveillance in public spaces for law enforcement, and AI systems designed to exploit psychological vulnerabilities.
“The supply chain implications are enormous,” said Luca Bertuzzi, who covers AI policy for POLITICO Europe. “If you’re a company that uses AI for recruiting, you’re responsible for ensuring that the AI vendor you bought the system from has complied with the Act. Most companies haven’t fully mapped their AI dependencies.”
General-purpose AI models — the large language models that power products like ChatGPT, Claude, and Gemini — have their own obligations under the Act. Models above a certain capability threshold must publish technical documentation and comply with EU copyright law.
Penalties for non-compliance can reach 35 million euros or 7 percent of global annual turnover, whichever is higher.
Leave a Reply