The Pulse
UK MPs Question OpenAI, Anthropic, Google and Meta on AI Safety
A Commons committee is due to question OpenAI, Anthropic, Google DeepMind and Meta on frontier-model safety and independent testing. The hearing will also examine whether the UK AI Security Institute can compel access to models or halt a re
AI.info Team · Written with AI ·
Testing access meets the power to stop a release
Britain’s AI Security Institute can assess a frontier model only if its developer lets it in—and the institute may not be able to compel that access or stop a launch. That gap sits at the centre of a Commons Business, Innovation, Science and Trade Committee session scheduled for Tuesday, 13 October, with representatives of Anthropic, OpenAI, Google DeepMind and Meta due to appear.
The committee plans to ask whether independent testing happens early enough to detect dangerous capabilities or misalignment, whether firms should face mandatory reporting duties, and who could halt a release if safety checks raise alarms. Its published agenda also names the choice between voluntary industry rules and mandatory regulation.
Committee chair Liam Byrne said: “So we need to get to the bottom of who is checking these systems, when those checks begin, and whether Britain’s safety experts have the power sound the alarm and stop the harm when things go wrong.”
Four lab representatives face the committee
The evidence is arranged in three sessions. Former OpenAI and Anthropic researchers Daniel Kokotajlo, Aengus Lynch, Steven Adler and Geoffrey Irving are listed for the 2 p.m. session. The committee then plans to hear from Anthropic’s Lord Booth-Smith, OpenAI’s Tom Duff Gordon, Google DeepMind’s Owen Larter and Meta’s Rob Sherman at 3 p.m.
At 4 p.m., witnesses are scheduled to include AI Security Institute director Henry De Zoete, Cabinet Office Permanent Secretary Cat Little and Alexandra Jones of the Department for Business, Innovation, Science and Trade. The session is listed as non-inquiry work, which Parliament says may address a topical issue but does not necessarily result in a report.
Ministers describe access as a work in progress
The committee’s concern about testing access comes as the government describes its own effort to maintain it. In a written statement to Parliament on 12 October, AI Minister Kanishka Narayan said the government had been working with AI labs and the US government to retain the institute’s access to the most powerful models before release.
Narayan said the institute’s more than 100 technical specialists were conducting evaluations of the latest generation of models, including Gemini 4 Argon. He also set out plans for an incident-response capability for serious cases involving AI agents and said the government was considering how the Cyber Security and Resilience Bill could better address AI-related cyber and loss-of-control risks.
The committee’s questions extend beyond testing
The hearing notice frames the discussion around the possible behaviour of rogue models and whether dangerous capabilities could advance faster than controls. It also asks whether the UK institute has sufficient authority to scrutinise models against the backdrop of what the committee describes as apparent US policy to halt pre-release of the most capable systems.
The committee has separately begun written questioning of AI Minister Kanishka Narayan about the government’s approach. The scheduled testimony from developers, former researchers and officials is set to put the central tension directly to the people on each side of it: whether access to independent testing can be assured, and what powers Britain has if a test identifies a serious risk.