EU Lawmakers Clinch Deal on Landmark Artificial Intelligence Legislation

Late Friday night, EU policymakers reached a political agreement on the final terms of the Artificial Intelligence Act. The bill was proposed in 2021 as a part of the EU’s broad digital legislation strategy, tackling algorithmic based harms with a risk based approach. The European Parliament and the European Council, the two legislative bodies in the EU, along with the European Commission, the EU’s executive branch, have spent the last 6 months negotiating the final terms based on each body’s draft of the bill.

The past month of negotiations stalled on two major issues:

• Law enforcement and national security exceptions, particularly in regard to the real time biometric surveillance prohibition
• How to regulate foundational models like ChatGPT

While there is no final text published, the Parliament, Council, and Commission have come to a compromise on the terms. According to the reporting, the final terms include a striking reduction of transparency requirements for law enforcement, the official relegation of real time biometric surveillance to the high risk category, a ban on predictive policing software based on personality, obligations for general purpose AI developers, and penalties up to 7% of a company’s annual global turnover.

The final step for the AI Act before it is officially adopted into law is for each body to formally approve the final terms in the third reading of the bill.

EPIC has long advocated for comprehensive privacy protections, rigorous testing protocols, expanded resources for evaluation of AI systems, and a government-whole effort to fighting algorithmic discrimination. Most recently, EPIC submitted comments to the Office of Management and Budget on the United States government’s procurement and use of AI. EPIC has also petitioned the FTC to commence a rulemaking on commercial use of artificial intelligence, and keeps track of state AI laws.