Microsoft's Brad Smith urged global AI rules and said Britain can help lead them.
Why it matters: Smith's intervention adds pressure on governments to establish baseline safeguards for advanced AI. It also positions Microsoft as a participant in shaping rules that could include mandatory testing, incident reporting and shutdown powers.
- Smith said AI is currently less regulated than a washing machine or carton of milk.
- He proposed independent evaluations, pre-release safety testing and mandatory disclosure of AI incidents.
- Smith said the UK is well positioned to lead international AI regulation because it is respected in the US, China and Europe.
- The UK plans to make frontier AI a central issue during its 2027 G20 presidency.
"Right now, AI is less regulated than your washing machine, it is less regulated than a carton of milk," Smith told ITV News.
He called for international cooperation, safeguards and monitoring rather than waiting for AI risks to become unmanageable. His proposed measures include independent evaluations and safety testing before launch, mandatory disclosure when AI agents go rogue or users are hacked, user notification after breaches, and emergency mechanisms that can take systems offline, according to the BBC.
Smith said Britain is particularly well positioned to help lead the effort because it is respected in the United States, China and Europe. "I think that this is the time when this government and the country can help the world address, with the right kind of balance and urgency, the safety that the public needs," he said.
The UK government has put frontier AI at the centre of its planned 2027 G20 presidency and says it will pursue global principles and standards for transparency, access and safe development, according to its stated plans.
But the country's approach remains decentralized. Existing sector regulators and point-of-use rules cover some AI applications, while promised binding rules for companies developing the most powerful models had not been introduced as of the House of Commons Library's June 10, 2026 briefing.
By the numbers
- 2027 - The year the UK plans to make frontier AI a central issue of its G20 presidency
- June 10, 2026 - Date of the House of Commons Library briefing that said binding frontier-model rules had not yet been introduced
- October 8, 2026 - Date of ITV's interview report featuring Smith's remarks
Yes, but: The milk comparison is a rhetorical shorthand, not a claim that AI has no regulation. Existing UK rules already cover some AI uses, and the UK's planned global principles have not been agreed by other G20 members.
What's next: The UK is expected to pursue global AI principles and standards during its 2027 G20 presidency, but the scope, enforcement and domestic implementation remain unresolved.