OpenAI, Anthropic Executives Testify on UK AI Oversight

OpenAI and Anthropic executives testified before a UK parliamentary committee on Oct. 13 as lawmakers examined whether advanced AI systems should face mandatory safety testing, incident reporting and release controls.

The Business, Innovation, Science and Trade Committee session began at 10:06 a.m. ET (14:06 UTC) and ended at 1:16 p.m. ET (17:16 UTC).

Lawmakers examined pre-release testing and reporting requirements for serious incidents involving powerful AI models. They also considered whether the UK AI Security Institute should have authority to require access to models or stop a release when safety concerns arise.

The hearing comes as the UK expands work on AI-related national security risks. The government is considering stronger legal protections through the Cyber Security and Resilience Bill regime, including reporting requirements for serious AI incidents and near misses.

Government pilots using frontier AI identified and addressed more than 1,000 potential cyber vulnerabilities across government systems. The UK AI Security Institute is also evaluating the latest generation of advanced models following work with developers and the U.S. government.

UK AI Minister Kanishka Narayan said, “We will legislate wherever necessary.”

MI5 Director General Sir Ken McCallum warned of “AI’s potential to change the stakes on existing or even new bioweapons.” MI5 launched an AI Threat Assessment Centre in 2026 with expertise from the UK AI Security Institute, the National Crime Agency and the National Cyber Security Centre.

OpenAI’s 2023 policy paper proposed pre-deployment risk assessments, external scrutiny, registration and reporting requirements, along with government enforcement mechanisms for frontier models. Anthropic’s public frontier-safety roadmap describes policy work aimed at helping policymakers manage AI risks across the industry.

The hearing follows the UK’s plan to provide the AI Security Institute access to advanced models for pre-release testing. The discussion puts voluntary cooperation between developers and government evaluators under scrutiny as officials consider whether legally enforceable requirements are needed for systems with potential cyber, terrorism and bioweapons applications.