The European Council has approved the AI Act, making it the first comprehensive law globally to regulate AI. The Act will be fully applicable after 24 months.

The Act regulates AI applications on a risk scale consisting of minimal, limited, high, and unacceptable levels. Unacceptable AI practices are prohibited, including unethical data collection and misuse for predictive policing.

Additionally, the Act also incorporates emergency procedures, exemptions, and the formation of an Office for Artificial Intelligence. "High-impact" AI models must fulfill transparency requirements before entering the market.

Furthermore, AI Regulation sandboxes are claimed to be established for AI innovation with certain "limited and specified" support and exemptions to smaller enterprises. Breaches of the Act will lead to either a percentage of the company's annual global turnover or a predetermined fine.

The new rules will come into force later this year. Specific provisions require generative AI models to be implemented within 12 months and most other requirements within two years.