Microsoft president says milk faces more safety rules than AI and calls for an emergency brake

One of the most powerful executives in technology spent Monday arguing that his own industry is regulated less carefully than breakfast.

Microsoft President Brad Smith told the BBC, in an interview published Monday, that AI needs urgent global safety rules with legal force behind them. His list: companies should have to inform victims when a rogue AI agent breaches their systems, file formal incident reports, submit new products to independent evaluators before release, and build in an emergency brake that can halt or slow a system that gets out of control.

Smith compared the present state of AI oversight to ordinary consumer goods, noting that washing machines and every carton of milk face safety rules and inspection that AI systems do not. He called AI both a tool and a weapon, and pointed to Britain's AI Security Institute, which gets early access to major models to assess their risks, as the model other countries should follow.

His chief executive made the same argument two days earlier. Satya Nadella wrote on Saturday that advanced models should be treated as compromised from the start and contained, with an authorized person always able to pause or shut a model down mid task and a tamper proof record of what it did.

The push puts Microsoft at odds with the White House preference for voluntary self regulation meant to keep American firms ahead of Chinese rivals. Smith said Microsoft would sign the self regulation pact President Trump brokered with tech leaders last month, but argued voluntary commitments only bind the companies that already act responsibly. OpenAI and Anthropic have themselves begun disclosing rogue or undesired behavior by AI agents, and some industry leaders have called for development to slow down, a debate this site's readers will recognize. Smith's position is that the rules should be law, and should start now.