OpenAI and Anthropic tell UK MPs voluntary AI pledges are not enough

Representatives of 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. WIRED describes this as deepening a public schism with Google DeepMind and Meta, both of which declined to back new legislation.

The session was held on Tuesday by the Business, Innovation, Science, and Trade Committee, a cross-party panel of MPs. It heard from four frontier AI firms, four former staff of the firms turned whistleblowers, and senior British civil servants.

Anthropic's head of Europe, Liam Booth-Smith, was asked whether the company's view was that AI safety now needs a statutory framework. "The rules should be binding," he said, adding that the company 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. He said "there should be some elements that should be mandatory," starting with incident reporting, and that "We don’t think basing everything on voluntary commitments for frontier risk is a good long-term plan."

Their rivals would not follow. Asked directly whether he supported a statutory framework, Google DeepMind's Owen Larter answered: "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."

Committee chair Liam Byrne told WIRED that exposing this divide in a second jurisdiction was the point. The committee set out "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."

A few hundred meters away, MI5 director Ken McCallum used his annual threat update to warn of bioweapon misuse as a significant threat from AI. He called it "not yet realized," but the misuse case to which he would attach greatest weight: AI's potential to change the stakes on existing or even new bioweapons. Byrne said the head of MI5 warning of dangerous capability uplifts for bioterrorists in one part of Whitehall, while the permanent secretary of the Cabinet Office is basically admitting the AI regulatory regime is voluntary in another, "tells you everything you need to know."

In the final part of the session, civil servants conceded 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. AISI director Henry de Zoete acknowledged the institute is entirely reliant on the cooperation of the companies and foreign governments to gain access, and that it has no ability to stop a company releasing a model AISI judged too dangerous. "What we’re not doing is kite marking," he said, using the British term for a safety certification. "We’re not saying yes or no to any sort of model." 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.

The frontier firms also made plain that British access runs through Washington. Duff Gordon said OpenAI provides its models predeployment to CAISI, AISI's US counterpart, and will keep providing them to AISI in the UK for as long as it is able to do so under the US government. Booth-Smith said Anthropic was unfortunately not able to provide full unrestricted access to Mythos 5.1 to UK AISI at that time, calling it a question for the US government, before adding that AISI now had access.

Four former lab researchers also gave evidence. Ex-OpenAI staffer Daniel Kokotajlo said he did not think the companies can be trusted to self-regulate. 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."

Afterwards Byrne said his committee would now dedicate a lot of time to 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, adding that the risks "are multiplying faster than our safety regime can control."

Key facts

OpenAI and Anthropic told the UK Business, Innovation, Science, and Trade Committee that voluntary pledges are not enough and new laws should be introduced; Google DeepMind and Meta declined to back new legislation.
Anthropic's Liam Booth-Smith said "The rules should be binding"; OpenAI's Tom Duff Gordon backed a statutory framework and an international convention, with incident reporting as the first mandatory element.
Civil servants conceded the regime is voluntary: AISI tests frontier models only when companies hand them over and cannot stop a release it judges too dangerous.
British access to models runs through Washington: OpenAI sends models to CAISI, and Anthropic said it could not give UK AISI full unrestricted access to Mythos 5.1 at that time.
Chair Liam Byrne wants ministers to use Britain's coming G20 presidency to broker an international convention that does not depend on the current White House.

Why it matters

Two of the leading frontier AI developers have now put on the British parliamentary record that voluntary commitments are not a good long-term plan and that binding rules are needed. Committee chair Liam Byrne said the aim was to expose the divide between firms in a second jurisdiction. The same session had civil servants admit that the UK's testing regime for frontier models depends entirely on company cooperation, which turns an abstract policy argument into an admission about how the system works today.

Who it affects

Frontier AI developers are the first group: OpenAI and Anthropic are on record favouring binding rules, while Google DeepMind and Meta pointed to standards and existing regulators. It also affects UK officials at AISI and the Cabinet Office, who described a testing role with no power to block a release. Beyond the UK, the hearing shows that access for British testers depends on the US government, so American policy shapes what UK institutions can see.

How to use it

This is a policy story rather than a product, so there is nothing to adopt. For readers tracking AI regulation, the concrete markers are the ones named at the hearing: incident reporting as the first element Duff Gordon said should be mandatory, a statutory framework, and an international convention that Byrne wants brokered through Britain's G20 presidency. The committee said it will spend a lot of time examining the UK's path forward as a middle power on AI.

How solid is it

The account is WIRED's report of an evidence session, with named witnesses and direct quotations, plus comments Byrne gave WIRED after the hearing. The positions of OpenAI and Anthropic are stated by their own representatives. Note that the source gives the day only as Tuesday and does not give a calendar date.

Risks and caveats

Nothing here is law. The source mentions no specific bill, legislative text or timetable, and does not say what the government plans to do in response. Larter's quoted answer was about focusing on standards; the source says only that DeepMind and Meta declined to back new legislation. Duff Gordon was described as more cautious than Anthropic. Byrne's comments about the Cabinet Office and the G20 are his own framing and ambition as committee chair.

“The rules should be binding”

— Liam Booth-Smith, Anthropic's head of Europe, to the UK parliamentary committee