Skip to content
NewsRegulation

Microsoft's Brad Smith says the UK, not Washington or Brussels, should write global AI safety rules

· by Pondero Newsdesk

The short version

Microsoft president Brad Smith told the BBC that consumer goods like milk and washing machines face more oversight than AI systems, and argued Britain's AI Security Institute makes the UK the right venue for binding global rules.

Microsoft's Brad Smith says the UK, not Washington or Brussels, should write global AI safety rules

Microsoft president Brad Smith argued on a BBC podcast, reported October 12, 2026, that milk and washing machines face tighter consumer-safety oversight than the AI systems being deployed into businesses and homes right now, and that the UK, not the US or EU, is positioned to fix that first.

What happened

Smith called AI a "weapon and a tool" that puts global cybersecurity at risk and said "the moment has arrived" for binding safety rules, per TechCrunch's report of the BBC interview. He pointed to Britain's AI Security Institute, launched in 2023, which gets early access to frontier models from US labs before they ship. That access has been politically shaky. The Trump administration reportedly asked OpenAI and Anthropic last month to withhold new models from UK testers while Washington ran its own review, per Politico. The AI Security Institute has since said it can still access and test the latest models, according to TechCrunch's reporting of that confirmation. Smith's read on why London still has an edge over Washington: "In Washington, D.C., people don't look at the U.K. and they don't look at London the way they look at Brussels," he told the BBC. "That's a strength."

Smith also took aim at the voluntary pledge tech leaders signed at the White House on October 4, in which OpenAI, Anthropic, Nvidia, AMD, SpaceX, Meta, and Google agreed to work with "independent auditors" on model safety and capability reviews. He called it a "good step" but said it does not go far enough, because "only the good companies" will actually comply. His specific asks: a legal duty to disclose when an AI system is breached by rogue agents, mandatory outside-evaluator sign-off before a model ships, and enforceable kill switches that can pause or slow a system if humans lose control of it.

The ask lands on labs already juggling two regulators

For frontier labs, Smith's proposal would mean a second mandatory review gate sitting alongside whatever Washington ultimately requires, not a replacement for one. If Smith's specific asks become policy, buyers deploying frontier models into UK-linked enterprise or government contracts should plan for an extra evaluator-approval step before release, and budget for a lag between US availability and UK or EU availability rather than assuming a simultaneous rollout. It also signals that Microsoft, which holds a deep commercial stake in OpenAI's models, is publicly pushing for more external scrutiny of the systems it sells, not less.

Smith's comments follow public calls from Anthropic's Dario Amodei and OpenAI's Sam Altman to slow AI development, after several incidents this year of AI agents acting outside intended bounds, per TechCrunch's reporting. That rival lab CEOs are already on record wanting a slower pace adds weight to Smith's push to make the voluntary October 4 audit pledge enforceable rather than aspirational.

What to watch next

The clearest signal to track is the next flagship release from OpenAI or Anthropic: whether the AI Security Institute gets pre-launch testing access the way it did for earlier models, or whether the White House's review delays that access again the way Politico reported happening in September. That outcome, not another statement from Smith, will show whether the UK's claimed edge is real policy weight or just rhetoric.

Sources