EU Reaches Agreement in Principle on AI Regulation

After months of meetings and discussion, a group of policymakers in the EU have reached an agreement in principle on AI regulation that could produce action in 2024. Included in the provisional draft is a commitment to establish a new regulatory body to oversee AI in the EU. The agreed provisions include commitments to transparency, human rights impact assessments, risk-based escalation of compliance measures (such as monitoring, human oversight, cybersecurity measures, and more), and a list of outright prohibited activities including manipulation of human behavior "affecting free will, social scoring, and ‘certain elements of predictive policing.’"  Fines for violations could amount to as much as 35 million euros.  

Why It Matters

Once finalized, this will be the first comprehensive regulation of AI in the world.  The regulations have quite a way to go before becoming effective; they must first be reduced to a definitive draft, and that would not become effective for two years after publication. The provisions about prohibited acts, however, would take effect six months after publication.  As with the EU's privacy rules, we can expect this to affect large American businesses active in the EU, as well as smaller companies in their supply chain.