Microsoft president says UK is best placed to drive AI regulation

Microsoft president Brad Smith told a BBC podcast that AI is both a weapon and a tool putting cybersecurity at risk, and that the moment has arrived for global safety rules. His sharpest line of contrast: everyday consumer products like milk and washing machines face more oversight than the companies building artificial intelligence.

Smith argues the United Kingdom is better placed than the United States or the European Union to lead. Britain’s AI Security Institute, launched in 2023, gets pre-deployment access to major U.S. frontier models and has become a leader in safety testing. That role has been contested—President Trump has warned against “globalist” testers and was reported last month to have asked Anthropic and OpenAI to withhold new models from British testers—yet AISI said Monday it can still access and test the latest models. Smith’s political read is blunt: Washington does not view London the way it views Brussels, and that is a strength.

We hear a Microsoft president trying to keep mandatory safety on the table after a White House “morally binding” self-audit pact that even he calls only a good first step. Voluntary rules, Smith says, bind the good companies and miss the ones you actually need to worry about. That is the right diagnosis of self-regulation theater.

Backing the UK is not neutral geopolitics. AISI’s pre-release testing access is one of the few institutional checks that still touches U.S. frontier labs from outside their own walls. Trump’s pressure to cut that access is a direct threat to that check. Smith’s claim that London looks less threatening than Brussels in D.C. is a lobbying map as much as a compliment—Microsoft wants rules, and it wants them written somewhere U.S. politics might tolerate.

The concrete asks are useful: a legal duty to alert an organization if rogue AI agents breach its systems; outside evaluator approval before release; kill switches to pause or slow systems if humans lose control. Those are the kinds of obligations that turn “safety culture” into enforceable practice. The risk is capture—big labs helping draft the rules they can already afford to meet, while smaller or less responsible actors either evade them or get locked out. Smith names that problem on the voluntary side; mandatory regimes can recreate it if compliance costs become the moat.

I’m watching whether AISI keeps pre-deployment model access under U.S. political pressure, and whether Microsoft’s calm-firm rhetoric shows up as support for binding rules rather than another round of pledges.

Smith’s comments arrive after tech leaders including Anthropic’s Dario Amodei and OpenAI’s Sam Altman called to slow AI development following incidents of rogue agents, and after a White House meeting where major AI and chip firms signed a morally binding pledge to work with independent auditors.