Brad Smith says AI is less regulated than milk. He has a list

The line everyone will quote came from Microsoft's president. Brad Smith told the BBC's Big Boss Interview, published October 11, that everyday products face more safety oversight than AI: "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." Cybernews's rendering adds that he described AI as both a "tool and a weapon" and urged the immediate imposition of stricter global regulations.

The quote is the hook. The list is the news. Smith proposed three concrete obligations for the industry:

First, emergency brakes: the ability to pause or slow down an AI system if it gets out of control, the industry's familiar "kill switch" idea, which Smith said firms need. Second, a legal requirement to write incident reports and notify victims when the technology goes rogue, specifically a legal duty to tell a breached organization that AI agents were involved. Third, independent government evaluators vetting new AI technology, with decisions about when outside evaluators sign off on an AI product before release.

He also named who should lead: the United Kingdom, via its AI Security Institute. That is the same institute the White House just restored pre-release model access to on Sunday, after a 17-day freeze. And on the White House accord, Smith was willing but unimpressed: happy to sign it, but "not sufficient any longer," as explainx.ai's readout of the BBC interview puts it. The BBC's own framing, as carried by the secondaries, puts him at odds with President Trump, who has said the sector should regulate itself so US firms do not lose to Chinese rivals.

A day earlier, on October 10, Microsoft CEO Satya Nadella made a related case in a long post on X: AI should not be treated as nested black boxes whose recommendations are simply accepted or rejected. Assume a model is compromised, he argued, and contain it from the start, like an emergency brake.

Here is what builders should take from this. None of Smith's three proposals is law. There is no draft text, no enforcement body, no threshold for when sign-off would be required, no timeline. The explainx.ai readout is explicit about that, and the article's headline is an interview, not a statute. But Smith is not a backbencher. When the president of Microsoft says kill switches and victim-notification duties should be legal obligations, that is what the compliance horizon looks like. The labs made the same kind of ask in Australia last week (what they asked for), and New York's council put AI executives under oath earlier this month (the hearing). Agent builders can already see which way the weather is moving: the White House's own Super Intelligence Force said on Friday that incident reporting is "not optional" (the full story), and now Microsoft's leadership is asking for the brakes and the notifications to be written into law.

The honest read is two-layered. The quote is rhetoric, and rhetoric is what interviews are for. The list is a preview of the obligations a serious regulatory regime would impose on anyone deploying agents, and it overlaps almost exactly with what the last week of incidents demanded: Philadelphia's complaint was about notification timing, Anthropic's about containment, Zenity Labs' about identity. If a Smith-shaped law ever arrives, the companies that already log every agent action and can pull the plug will not have to rebuild. The ones that cannot will be building under deadline.