Microsoft president Brad Smith handed the BBC’s podcast a line on Monday that is easy to remember: “Right now, your washing machine is subject to more safety rules than AI. Every carton of milk that you see in the grocery store is subject to safety rules and inspection.” He calls AI a tool and a weapon at once, and then: “Let’s get started today.”
Four demands, all specific
Smith wants global safety rules, and he spells out what they should contain. Emergency brakes first, the thing the industry likes to call a kill switch, to pause or slow a model that gets out of control. Second, a legal requirement to file incident reports. Third, an obligation to tell a victim when an agent breaks loose and ends up inside their system. Fourth, independent government evaluators who sign off on new AI products before release.
None of that is a new note, just the same one struck last week from elsewhere. I wrote about Satya Nadella’s emergency brake for AI models on Sunday, and mandatory incident reporting has been in the Super Intelligence Force order since Friday. So Smith is asking for legislation that mirrors what the US just created as an obligation — only worldwide.
Why London rather than Brussels
His case for the UK rests on the AI Security Institute, launched in 2023 under Rishi Sunak as the AI Safety Institute. “Name the country that is the leader in having this technical capability,” Smith says of people in tech, “they would say, it’s very simple. It’s the United Kingdom. They are the best.”
There’s a political argument on top. Smith thinks Washington would accept British standards more readily than European ones: “In Washington DC, people don’t look at the UK and they don’t look at London the way they look at Brussels — that’s a strength.”
The distance to his own president
Donald Trump says the sector should regulate itself so US firms don’t lose ground to Chinese rivals. Anthropic, OpenAI and Google DeepMind have all backed the idea of slowing AI development down; Trump has pushed against it. Smith places himself visibly on the other side, but phrases it carefully: he doesn’t believe the US government is as opposed as Trump’s comments make it sound.
On the “morally binding” document the heads of OpenAI, Anthropic, Nvidia, SpaceX, Meta and Google signed at the White House in late September, he says: good step, necessary step, Microsoft is happy to sign it. “But I don’t think it’s sufficient any longer.”
Self-regulation fails on the wrong companies
Smith’s strongest argument comes last, and it isn’t a moral one. The trouble with any self-regulatory regime, he says, is that “the good companies will do what you want.” It’s the less responsible ones you have to worry about.
That lands close to everything else this week. A lab that harvests safeguards instead of building them won’t be held by a signature. Which makes the milk comparison more than a headline. Food inspection works precisely because nobody opts into it.