Microsoft urges EU to reassess AI Act as models evolve

Microsoft President Brad Smith has urged European officials to reconsider whether the EU’s AI Act remains suited to evolving AI technology. European Commission officials maintain that the law, drafted in 2023, is strong enough to address security risks posed by the latest models and can accommodate AI agents.

Google DeepMind co-founder Demis Hassabis separately warned that older laws may fail to address current problems. He did not specifically mention the AI Act, distinguishing his broader concerns about regulation from Smith’s call to reassess the European framework.

An October paper from Hungary raised concerns about models tested before release, authorities’ access to training information and liability for harm caused by agents. Those issues highlight questions surrounding oversight and accountability as European officials defend the law’s ability to handle newer AI systems.