OpenAI and Anthropic Tell UK MPs Voluntary AI Safety Pledges Are Not Enough; Google DeepMind and Meta Decline to Back New Laws
At a Commons committee hearing on Oct. 13, Anthropic said there should be binding rules on AI companies and OpenAI backed a statutory framework with some mandatory elements, starting with incident reporting, WIRED reports. Google DeepMind and Meta did not back new legislation. This was an evidence session; officials described the UK's current frontier-AI safety regime as voluntary.
Correction, October 14, 2026: An earlier version made overly broad statements that no AI bill had been tabled or legislation proposed. Parliament lists an Artificial Superintelligence Bill. This report concerns testimony at an evidence session and officials’ description of the current frontier-AI safety regime.
OpenAI and Anthropic told British MPs on Tuesday, October 13 that voluntary safety pledges are no longer enough and that new laws should be introduced, while Google DeepMind and Meta declined to back new legislation, WIRED reports. The positions were given as evidence to the Commons Business, Innovation, Science and Trade Committee. These were positions presented at an evidence session; officials told the committee the UK's current frontier-AI safety regime is voluntary.
What each company said#
"The rules should be binding," Anthropic's Lord Booth-Smith told the committee when asked whether AI safety needs a statutory framework, according to WIRED. He said the company has "consistently believed that there should be binding rules on companies and that voluntary commitments are not enough," WIRED reports; the committee's automated session transcript carries the same line.
OpenAI's Tom Duff Gordon was more cautious, WIRED reports, 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."
Asked whether he supported a statutory framework, Google DeepMind's Owen Larter said, "Focus on the standards is what we would be doing," according to WIRED. Meta's Rob Sherman pointed to Britain's existing regulators, saying they were "not writing on a blank page."
What officials said about the current regime#
The UK's AI Security Institute (AISI) tests frontier models before release only when companies hand them over, WIRED reports. AISI director Henry de Zoete said the institute relies on the cooperation of companies and foreign governments for access, and acknowledged it could not stop a release it judged too dangerous. Asked whether the regime was voluntary, Cabinet Office permanent secretary Cat Little said: "At the moment, yes, it is."
OpenAI on incidents and Astra 6.1#
Duff Gordon told the committee that OpenAI paused reinforcement-learning training on its most advanced model after the July Hugging Face incident and held back a planned release, Astra 6.1, Courthouse News Service reports. He said he could not assure the committee that no further incidents of agents accessing government websites would emerge: "There are a handful of others that I am aware of that we have notified, and there may be more to come." The reason for holding Astra 6.1 is OpenAI's own account.
Elsewhere the same day#
MI5 director general Ken McCallum used his annual threat update to say that the misuse case he would "attach greatest weight" to is "AI's potential to change the stakes on existing or even new bioweapons," WIRED reports. Former OpenAI researcher Daniel Kokotajlo told MPs, "I don't think the companies can be trusted to self-regulate," according to WIRED.
What comes next#
The committee said it would produce a special report on the hearing ahead of possible government amendments to the Cyber Security Bill, and chair Liam Byrne announced a new inquiry into AI safety, Courthouse News reports.
Limits#
- The hearing was an evidence session, not a vote enacting binding rules.
- The official committee page confirms the oral-evidence session. Quotes come from WIRED and Courthouse News; the automated transcript on Parallel Parliament was used only to check wording.