OpenAI and Anthropic Back Binding Safety Rules at UK Parliamentary Hearing

Representatives from OpenAI and Anthropic told a British parliamentary hearing on Tuesday that voluntary industry commitments are no longer sufficient to govern frontier artificial intelligence, backing statutory rules and creating a public divide with Google DeepMind and Meta.

Testifying before the House of Commons Business, Innovation, Science, and Trade Committee, Anthropic head of Europe Liam Booth-Smith said the company has long maintained that rules should be binding. OpenAI vice president Tom Duff Gordon supported a statutory framework and an international convention, arguing that voluntary regimes are inadequate for frontier risk and calling for mandatory incident reporting. In contrast, Google DeepMind representative Owen Larter urged lawmakers to focus on standards rather than legislation, while Meta's Rob Sherman pointed to existing British regulators, as reported by Wired .

The hearing highlighted the constraints facing Britain's current oversight model. Henry de Zoete, director of the UK's AI Security Institute (AISI), and Cat Little, the Cabinet Office's top civil servant, acknowledged that AISI testing relies entirely on voluntary cooperation from companies. De Zoete stated that the institute has no authority to block releases or issue formal safety certifications, noting that British testing access depends heavily on US government approvals.

Parallel national security warnings underscored the debate. Nearby in Whitehall, MI5 director Ken McCallum warned in an annual threat briefing that artificial intelligence could lower barriers to developing existing or novel biological weapons. Committee chair Liam Byrne told Wired that the committee aims to use the UK's upcoming G20 presidency alongside South Korea to pursue an international convention, arguing that the risks of advanced models are outpacing voluntary safety measures.