UK Parliament AI Safety Hearing: OpenAI and Anthropic Call for Binding Rules as Google and Meta Decline

What happened

On Tuesday, Oct 13, 2026, the UK House of Commons Business, Innovation, Science and Trade Committee held an evidence session on AI safety. According to WIRED, those giving evidence included representatives of four frontier AI companies (OpenAI, Anthropic, Google DeepMind and Meta), four former employees who became whistleblowers, and senior British civil servants.

Anthropic's head of Europe, Liam Booth-Smith, was asked whether AI safety now needs a statutory framework, and said the company has consistently believed there should be binding rules on companies, and that voluntary commitments are not enough.

The rules should be binding.

— Liam Booth-Smith, Anthropic's head of Europe

OpenAI vice president Tom Duff Gordon took a more cautious but supportive position: he backed a statutory framework and an international convention, saying the starting point should be mandatory incident reporting.

We don't think basing everything on voluntary commitments for frontier risk is a good long-term plan.

— Tom Duff Gordon, OpenAI vice president

Their rivals would not follow. Asked directly whether he supported a statutory framework, Google DeepMind's Owen Larter answered that what his company would do is "focus on the standards"; Meta's Rob Sherman pointed to Britain's existing regulators, telling MPs they were "not writing on a blank page."

Committee chair Liam Byrne told WIRED after the hearing that the point of the session was "to light up the wide differences in opinions between some of the wealthiest firms on Earth" and to show "that you couldn't rely on voluntary self-regulation."

On the same day as the hearing, MI5 director Ken McCallum used his annual threat update to warn that AI could change the stakes on existing or even new bioweapons.

Key facts

1. The hearing: On Oct 13, 2026, the UK House of Commons Business, Innovation, Science and Trade Committee held an AI safety evidence session, with representatives of OpenAI, Anthropic, Google DeepMind and Meta, four whistleblower ex-employees and senior civil servants
2. Pro-legislation camp: Anthropic's Liam Booth-Smith called for binding rules; OpenAI's Tom Duff Gordon backed a statutory framework and an international convention, starting with mandatory incident reporting
3. The other camp: Google DeepMind's Owen Larter said to "focus on the standards"; Meta's Rob Sherman pointed to Britain's existing regulators and said MPs were "not writing on a blank page"
4. MI5 warning the same day: MI5 director Ken McCallum warned in his annual threat update that AI could change the stakes on existing or even new bioweapons; on the same day MI5 announced an AI threat assessment centre (AITAC) that will work with the NCSC, the NCA and AI Safety Institute experts on all-source AI risk assessment
5. Where AISI stands: Britain's AI Safety Institute can currently test frontier models before release only when companies voluntarily hand them over, and has no power to block releases it deems unsafe; institute director Henry de Zoete has acknowledged the system depends entirely on company cooperation

Background

Two days before the hearing, the US and the UK had just gone through a dispute over model testing access. On Sept 24, the US asked developers such as OpenAI and Anthropic to pause sharing new models with British testers until US national security reviews were complete; the restriction lasted about 17 days, during which Anthropic's Claude Mythos 5.1 was opened only to US institutions. According to Reuters relaying a Telegraph report on Oct 11, the US reversed the decision and Britain's AI Safety Institute regained pre-release access. Related coverage

On the same day as the hearing, around 400 scientists from 34 countries published an open letter calling for an international body to set containment standards and handle incident reporting for advanced cyber-capable AI systems. Signatories included Turing laureates Geoffrey Hinton and Yoshua Bengio. The letter proposed air-gapping frontier models from external networks with independent oversight, and a registry of significant security incidents eventually made public, arguing voluntary industry commitments are "not enough". It came about a month after the White House voluntary accord was signed. Related coverage

US policy was tightening as well. On Sept 29, Anthropic, OpenAI, Google, Meta, xAI and Nvidia signed the White House voluntary accord committing to voluntary safety measures; on Oct 11, the Trump administration demanded mandatory security-incident reporting from AI companies, with its new AI task force making its debut. Meanwhile, on Oct 9, EU tech chief Henna Virkkunen said the bloc's AI Act is sufficient to address risks from rogue AI.