OpenAI and Anthropic Ask UK to Regulate Them — Rivals Refuse

In an unusual move, OpenAI and Anthropic appeared before a British parliamentary committee and asked lawmakers to pass binding rules for powerful AI systems — while their rivals Google DeepMind and Meta refused to back new legislation.

A Divided Industry

The hearing, held by the Business, Innovation, Science, and Trade Committee on Tuesday, exposed a deep split among the world's most powerful AI companies over how to keep dangerous technology in check.

Liam Booth-Smith, Anthropic's head of Europe, was direct. "The rules should be binding," he told MPs, adding that the company has "consistently believed that there should be binding rules on companies and that voluntary commitments are not enough."

Tom Duff Gordon, OpenAI's vice president for Europe, the Middle East and Africa, was more cautious but still supported mandatory requirements, especially for incident reporting. "We don't think basing everything on voluntary commitments for frontier risk is a good long-term plan," he said.

The other two companies pushed back. Google DeepMind's Owen Larter told the committee to "focus on the standards" instead of new laws. Meta's Rob Sherman pointed to existing British regulators, saying the UK was "not writing on a blank page."

The Security Warning

While the executives debated regulation, a security warning came from just a few hundred meters away. MI5 Director General Ken McCallum delivered his annual threat update, warning that AI could be misused to develop bioweapons — a danger he described as a "significant threat."

He also announced a new AI Threat Assessment Centre within the intelligence community.

Committee chair Liam Byrne highlighted the contradiction: "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."

A Voluntary System With No Power

British civil servants confirmed the system's limits. Henry de Zoete, director of the UK's AI Security Institute (AISI), acknowledged the institute can only test models when companies voluntarily hand them over. When asked what would happen if a model was deemed too dangerous but released anyway, he said AISI had no power to stop it.

OpenAI's Duff Gordon revealed that the company shares its models with AISI's US counterpart first, and access to the UK body depends on US government cooperation.

Former OpenAI researcher Daniel Kokotajlo told MPs: "I don't think the companies can be trusted to self-regulate." Another ex-staffer, Steven Adler, said he was "not confident in the safety culture across the industry."

Byrne told WIRED the committee would now focus on how Britain, as a middle power, can help broker an international AI safety convention through its upcoming G20 presidency.

"The risks," he said, "are multiplying faster than our safety regime can control."

Source markers

¹ Voluntary commitments not enough, says Anthropic's Booth-Smith . MI5 warns of bioweapon misuse . AISI has no power to prevent dangerous model releases . ² McCallum announces AI Threat Assessment Centre . ³ Former staff criticize industry self-regulation .

Key facts

OpenAI and Anthropic testified before the UK parliament calling for binding AI regulations
Google DeepMind and Meta refused to support new legislation, preferring voluntary standards
MI5's director warned AI could be used to develop bioweapons
The UK's AI Security Institute can only test models when companies volunteer them
AISI has no power to stop companies from releasing dangerous models