Anthropic, OpenAI Back UK AI Safety Laws at Hearing — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic head of Europe Liam Booth-Smith told the committee "the rules should be binding," saying 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 backed a statutory framework and international convention, with mandatory elements starting with incident reporting, but was more cautious than Anthropic in his language.
- Google DeepMind's Owen Larter declined to support new legislation, telling MPs "focus on the standards is what we would be doing," while Meta's Rob Sherman pointed to Britain's existing regulators as not "writing on a blank page."
- AISI director Henry de Zoete acknowledged the institute is "entirely reliant" on company cooperation to access frontier models and has no power to prevent release of a model it deems too dangerous — he told the committee AISI is "not kite marking."
- Cat Little, permanent secretary at the Cabinet Office which houses AISI, conceded under questioning that the current UK AI regime is voluntary: "At the moment, yes, it is."
- Anthropic told MPs it was "not able to provide full unrestricted access" to its Mythos 5.1 model to UK AISI — "that's a question for the US government" — though access was eventually granted.
- MI5 director Ken McCallum, speaking at a separate address a few hundred meters away, said the AI misuse case he'd attach greatest weight to is bioweapons: "not yet realized, but ... AI's potential to change the stakes on existing or even new bioweapons."
Why it matters: Britain's AI safety testing has no legal teeth — AISI's own director admitted he cannot block deployment of a model he judges dangerous, and frontier model access to UK regulators is gated by Washington, not Whitehall. That structural dependency is why committee chair Liam Byrne said he wants ministers to use Britain's G20 presidency to broker a convention that does not depend on the current White House, putting UK policy on a collision course with American AI oversight.
Ask SkimNews
