OpenAI and Anthropic tell UK MPs: binding AI safety rules, not only voluntary pledges

Representatives of OpenAI and Anthropic told a UK parliamentary committee on 13 October 2026 that voluntary AI safety pledges are no longer enough and that binding rules are needed — a public split with Google DeepMind and Meta, which declined to endorse new primary legislation, according to WIRED.

What was said

Anthropic’s Europe head Liam Booth-Smith told the Business, Innovation, Science and Trade Committee: “The rules should be binding,” arguing voluntary commitments “are not enough.” OpenAI’s Tom Duff Gordon backed a statutory framework and international convention, saying at least some elements — starting with incident reporting — should be mandatory.

Google DeepMind’s Owen Larter said the focus should be on standards. Meta’s Rob Sherman pointed to Britain’s existing regulators, saying MPs were “not writing on a blank page.”

AISI access is still opt-in

UK AI Security Institute director Henry de Zoete and Cabinet Office permanent secretary Cat Little told MPs the regime is currently voluntary: AISI tests frontier models only when labs choose to share them, and has no power to block a release. Lab executives noted UK pre-deployment access often runs through US arrangements (CAISI).

Context

Whistleblowers including Daniel Kokotajlo urged MPs not to trust self-regulation. Separately the same day, MI5 director Ken McCallum warned of AI’s potential to raise bioweapon stakes. Committee chair Liam Byrne said the hearing was meant to show industry disagreement and that voluntary self-regulation cannot be relied on alone.