Representatives for OpenAI and Anthropic told a British parliamentary hearing that voluntary pledges were no longer enough to ensure safety and that new laws should be introduced, deepening a public schism with Google DeepMind and Meta, both of which declined to back new legislation.
The Business, Innovation, Science, and Trade Committee, comprising a cross-party panel of MPs, held an evidence session on Tuesday with appearances from four frontier AI firms, four former staff of the firms turned whistleblowers and senior British civil servants.
“The rules should be binding,” Anthropic’s head of Europe Liam Booth-Smith told the committee, when asked if the company’s view was that AI safety now needs a statutory framework. The company, he said, has “consistently believed that there should be binding rules on companies and that voluntary commitments are not enough.”
OpenAI vice president Tom Duff Gordon was more cautious but still supportive, backing a statutory framework and an international convention. Duff Gordon said “there should be some elements that should be mandatory,” starting with incident reporting. “We don’t think basing everything on voluntary commitments for frontier risk is a good long-term plan,” he said.
However, their rivals would not follow. Asked directly whether he supported a statutory framework, Google DeepMind’s Owen Larter said: “Focus on the standards is what we would be doing.” Meta’s Rob Sherman pointed to Britain’s existing regulators, telling MPs they were “not writing on a blank page.”
The hearing exposed an existing divide on the subject of regulation, putting the conflicting approaches of AI leaders on the British parliamentary record. Exposing this divide in a second jurisdiction was the point, according to select committee chair Liam Byrne. The committee set out “to light up the wide differences in opinions between some of the wealthiest firms on Earth,” he told WIRED after the hearing, and to show “that you couldn’t rely on voluntary self-regulation.”
While the four executives gave evidence to the committee on AI safety, only a few hundred meters away, Britain’s security service director issued a warning about the dangers of AI.
MI5 director Ken McCallum used his annual threat update to warn of bioweapon misuse as a significant threat from AI. “Not yet realized, but the misuse case to which I’d attach greatest weight, is AI’s potential to change the stakes on existing or even new bioweapons,” McCallum said.
Responding to this comment, Byrne told WIRED: “The fact that the head of MI5 is warning about incredibly dangerous capability uplifts for bioterrorists in one part of Whitehall, and then in another part of Whitehall the permanent secretary of the Cabinet Office is basically admitting that our regulatory regime for AI is voluntary, tells you everything you need to know.”
In the final part of the evidence session, civil servants conceded to the committee how little the current system compels. Britain’s AI Security Institute (AISI) tests frontier models before release, but only when companies choose to hand them over.
Henry de Zoete, director of AISI, acknowledged that the institute was entirely reliant on the cooperation of the companies and foreign governments in order to gain access. Cat Little, the top civil servant at the Cabinet Office, which houses AISI, was pressed on whether the whole regime was voluntary. “At the moment, yes, it is,” she said.
When asked what would happen if AISI judged a model to be too dangerous, but it was released by a company anyway, de Zoete acknowledged AISI had no ability to prevent them from doing so.
“What we’re not doing is kite marking,” de Zoete said, using the British term for a safety certification. “We’re not saying yes or no to any sort of model.” He added: “These companies are either American or Chinese who are building at the frontier … There’s only so much we, as the UK, can do in that kind of context.”
The frontier executives also made plain that British access runs through Washington. “We provide our models predeployment to CAISI,” OpenAI’s Duff Gordon said, referring to AISI’s US counterpart. “We will continue to provide our models pre-deployment to AISI in the UK … and as long as we’re able to do that under the US government, we will continue to do it.”
The tension of the existing arrangement was made clear to British officials by Anthropic’s Booth-Smith. “We unfortunately were not able to provide full unrestricted access to [Mythos 5.1] to UK AISI at that time … That’s a question for the US government,” he said before adding that AISI now had access.
The session also heard from four former lab researchers, among them ex-OpenAI staffer Daniel Kokotajlo. “I don’t think the companies can be trusted to self-regulate,” he told MPs. Steven Adler, who left OpenAI in late 2024, said he was “not confident in the safety culture across the industry.” Geoffrey Irving said the labs believe they are locked in a race that governments will not interrupt and that they cannot leave. “Both of those are false,” he said. “They could just stop.”
Following the hearing Byrne told WIRED that his committee would now dedicate a lot of time examining the UK’s path forward as a middle power on AI. He wants ministers to use Britain’s coming G20 presidency, alongside South Korea, which follows it in the chair, to broker an international convention that does not depend on the current White House. “The G20 presidency will outlive the US presidency of Trump,” he said.
“The risks,” he added, “are multiplying faster than our safety regime can control.”
