The final agreement subjects GPAI to a two-tier system, allowing some flexibility for companies. The act also bans systems with an "unacceptable" risk level, and while a total ban on facial recognition was initially proposed, the approved draft includes exceptions for limited use in law enforcement. The full text of the AI Act will not be available for several weeks, and the legislation is expected to become law by mid-2024, with all provisions coming into force gradually over the next two years.
Key takeaways:
- European lawmakers have reached a provisional deal on the EU AI Act, which aims to regulate artificial intelligence technologies. The Act focuses on high-risk AI systems and general-purpose AI systems (GPAI), such as OpenAI's GPT models.
- France, Germany, and Italy initially resisted the Act, fearing that strict regulations could stifle innovation and harm startups. However, the final agreement includes a two-tier system that allows some flexibility for companies.
- The Act also includes provisions for systems with an "unacceptable" risk level, which are effectively banned. This includes certain high-tech surveillance tools, although the specifics are still being debated.
- The final legislation is still subject to change and is expected to become law by mid-2024, with all provisions coming into force gradually over the next two years. This gives AI companies time to ensure their products and services will be compliant with the rules.