OpenAI and Anthropic Tell British Lawmakers "Regulate Us"; Google DeepMind and Meta Decline to Back New Law

On October 13, the UK Parliament's Business, Innovation, Science, and Trade Committee held an evidence session. WIRED reports that representatives of OpenAI and Anthropic told MPs that voluntary pledges are no longer enough for safety and that new laws should be introduced. This put an open split with Google DeepMind and Meta on the British parliamentary record, since both declined to back new legislation. The witnesses were four frontier AI firms, four former staff of those firms who became whistleblowers, and senior civil servants. Committee chair Liam Byrne told WIRED afterward that the aim 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."

The four positions were clear. Anthropic's head of Europe, Liam Booth-Smith, said: "The rules should be binding." He said the company has consistently believed that binding rules are needed and that voluntary commitments are not enough. OpenAI vice president Tom Duff Gordon was more careful, but he also backed a statutory framework and an international convention. "There should be some elements that should be mandatory," he said, starting with incident reporting, and "We don't think basing everything on voluntary commitments for frontier risk is a good long-term plan." Asked directly whether he backed a statutory framework, Google DeepMind's Owen Larter said only: "Focus on the standards is what we would be doing." Meta's Rob Sherman told MPs that Britain's existing regulators are in place and "we're not writing on a blank page." On the same day, MI5 director Ken McCallum used his annual threat update to warn that AI could change the stakes of bioweapons, which he called the misuse case "to which I'd attach greatest weight."

The last part of the session showed how soft the current system is. The AI Security Institute tests frontier models before release, but only when companies choose to hand them over. Director Henry de Zoete admitted that it relies entirely on the cooperation of companies and foreign governments, and that if it judged a model too dangerous and the company released it anyway, it could not stop that. "What we're not doing is kite marking," he said. "We're not saying yes or no to any sort of model." He added that the frontier builders are American or Chinese, so there is only so much the UK can do. Cat Little, the top civil servant at the Cabinet Office, was asked if the whole regime is voluntary. "At the moment, yes, it is," she said. Access also runs through Washington. OpenAI's Duff Gordon said the firm gives models to the US CAISI and the UK AISI before deployment, "and as long as we're able to do that under the US government, we will continue to do it." Anthropic's Booth-Smith said the company "was not able to provide full unrestricted access" to Mythos 5.1 to the UK institute at first, "a question for the US government", but that AISI now has access.

The four former lab researchers were more direct. Daniel Kokotajlo, once at OpenAI, said he does not think the companies can be trusted to self-regulate. Steven Adler, who left OpenAI in late 2024, said he is 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." After the hearing, Byrne said his committee will now spend much time on the UK's path as a middle power in AI. He wants ministers to use Britain's coming G20 presidency, with South Korea next 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, and "The risks are multiplying faster than our safety regime can control."

Two cautions for readers. First, this was a hearing, not a law. No bill passed, and no company promised to change what it does; the witnesses stated positions. Second, the firms that ask for binding rules also keep building and releasing frontier models, so asking for regulation does not mean slowing down. Critics often ask whether such calls can help firms shape rules in their own favor. The WIRED report does not answer that, but it puts each side's own words on the record, and that is useful for the legislative debate that follows.