Senior staff from four leading AI companies gave evidence to MPs on Tuesday 13 October 2026, and split on whether AI safety should be required by law. Anthropic and OpenAI said at least some rules should be binding. Google DeepMind and Meta did not back a new law. OpenAI also told MPs it had scrapped the release of a model, GPT-6.1 Astra, over safety concerns.
The hearing was held by the House of Commons Business, Innovation, Science and Trade Committee. It also heard from four former AI lab researchers and from senior civil servants.
Where the companies stood
According to WIRED’s report of the session, Anthropic’s Liam Booth-Smith, who leads the company’s work in Europe, was asked whether AI safety needs a legal framework. “The rules should be binding,” he said. He added that Anthropic has “consistently believed that there should be binding rules on companies and that voluntary commitments are not enough.”
OpenAI’s vice president Tom Duff Gordon was more cautious, WIRED reported. He said “there should be some elements that should be mandatory”, starting with incident reporting, meaning firms would have to tell authorities when something goes wrong. “We don’t think basing everything on voluntary commitments for frontier risk is a good long-term plan,” he said.
Asked directly whether he supported a legal framework, Google DeepMind’s Owen Larter said, according to WIRED: “Focus on the standards is what we would be doing.” Meta’s Rob Sherman pointed to the regulators the UK already has, telling MPs they were “not writing on a blank page.”
OpenAI says it held back GPT-6.1 Astra
The Independent reported that Duff Gordon told MPs OpenAI had scrapped the release of GPT-6.1 Astra over safety concerns. “We found it didn’t refuse enough, it hadn’t progressed enough with alignment,” he said. A model that does not “refuse enough” is one that goes along with harmful requests it should turn down. “Alignment” means making an AI system reliably do what its makers and users intend.
“It is evidence that OpenAI takes safety really seriously, that we are not racing to the bottom,” he added.
The UK’s testing is voluntary
Britain’s AI Security Institute (AISI) tests powerful models before release, but only when companies choose to hand them over. WIRED reported that AISI’s director, Henry de Zoete, accepted that the institute could not stop a company releasing a model that it judged too dangerous.
“What we’re not doing is kite marking,” he said, referring to the British safety-approval mark. “We’re not saying yes or no to any sort of model.”
Cat Little, the most senior civil servant at the Cabinet Office, which houses AISI, was asked whether the whole system was voluntary. “At the moment, yes, it is,” she said.
Former lab staff were blunter
Former OpenAI researcher Daniel Kokotajlo told MPs: “I don’t think the companies can be trusted to self-regulate,” WIRED reported. Steven Adler, who left OpenAI in late 2024, said he was “not confident in the safety culture across the industry.” Geoffrey Irving said labs believe they are stuck in a race that governments will not stop and that they cannot leave. “Both of those are false,” he said. “They could just stop.”
What comes next
Committee chair Liam Byrne told WIRED after the hearing that he wants ministers to use the UK’s coming G20 presidency to broker an international agreement on AI safety that does not depend on the current US administration.
What this means
- The big labs are divided. Two of the four firms backed some binding rules; two did not.
- The UK has no power to block a model. Its safety testing depends on companies agreeing to take part.
- The Astra decision is OpenAI’s own account. OpenAI presented the scrapped release to MPs as proof that it takes safety seriously.
Sources
- WIRED, “OpenAI and Anthropic Tell British Lawmakers: Regulate Us”, 13 October 2026. https://www.wired.com/story/openai-and-anthropic-tell-british-lawmakers-regulate-us/
- The Independent (Henry Saker-Clark), “OpenAI ‘not racing to the bottom’ on AI safety, MPs told”, 13 October 2026. https://www.independent.co.uk/news/uk/politics/mps-openai-chatgpt-anthropic-claude-b3066114.html
Related stories
- Anthropic and OpenAI Sought Significant Control Over What Outside AI Evaluators Can Publish, Reuters Reports
- White House Reportedly Reverses Course, Restoring UK AI Security Institute’s Early Access to US AI Models
- UK AI Security Institute Sets Out Its Priorities for Technical Best Practice in Frontier AI Development
- Hundreds of Scientists, Including Geoffrey Hinton and Yoshua Bengio, Call for an International Body on AI Containment and Incident Reporting
- Nearly Astra, at a Fifth of the Price: What OpenAI’s GPT-6.1 Sol Means for Builders
- MI5 Chief Says AI-Enabled Bioweapons Are His Biggest AI Worry, and Launches an AI Threat Assessment Centre
