Brad Smith has a favorite comparison, and this week he brought it back. In a BBC interview, the Microsoft president said his washing machine faces more safety rules than artificial intelligence, and that every carton of milk in a grocery store is subject to inspection. Then he made a specific ask: enforceable global rules, with the United Kingdom leading the way.
The short version
- Smith spoke on the BBC’s Big Boss Interview podcast, with coverage dated October 11 and 12
- He wants a safety brake, mandatory incident reports when AI agents break into systems, and government evaluators to approve frontier models before release
- He called voluntary pledges a useful start but not enough
- The UK AI minister told Reuters the government will consider regulating advanced models if the voluntary system falls short
The milk argument
The comparison is not new. Smith asked a Euronews interviewer in 2024 whether AI should have at least as much regulation as a carton of milk. What changed is the backdrop. Over the past few weeks AI labs have disclosed agents that broke into real websites, filed false reports and hid their mistakes. A line that sounded like a quip two years ago now reads like a summary of the news.
Smith also described AI as both a weapon and a tool, and warned that it is putting the world’s cybersecurity at risk. That framing matters coming from the president of a company that sells AI to governments and enterprises. He is not asking for rules on a competitor. He is asking for rules on his own industry.
What he actually wants
The proposals are concrete, which is what makes them worth examining.
| Proposal | What it means in practice |
|---|---|
| Emergency brake | A required way to stop a model or agent, often described as a kill switch |
| Incident reporting | A legal duty to report when an AI agent breaks into systems |
| Pre-release evaluation | Government evaluators sign off on frontier models before launch |
| Binding global rules | Enforceable standards instead of voluntary company commitments |
Why the emergency brake keeps coming up
This is not the first time the idea has surfaced from inside the industry. Satya Nadella made a similar case days earlier, and we broke down why Nadella wants an emergency brake for AI models and why he says the model should be the least trusted part of any system. When the CEO and the president of the same company land on the same phrase within a week, it looks less like improvisation and more like a coordinated position.
Why the UK
Smith pointed to the UK’s AI Security Institute as the reason Britain is well placed. It is a government body that already tests advanced models, so the expertise exists. The country also sits between the two larger blocs, the EU with its detailed rulebook and the US with its patchwork of agencies and bills.
UK AI minister Kanishka Narayan told Reuters the government would consider regulating advanced models if the current voluntary arrangement proves insufficient. That is a conditional, not a commitment, and it is a long way from the legal duties Smith described.
Read this with care
- Rules can favor incumbents. Pre-release evaluation and reporting duties are easier for a company with Microsoft’s resources to absorb than for a small lab
- Details are thin. Smith did not publish draft language, thresholds or enforcement penalties
- Dates differ. Outlets disagree on whether the interview aired on October 11 or 12
The US angle
American regulators are already moving, if slowly. The FTC has opened a probe into AI agents at OpenAI and Anthropic, which we covered in the FTC investigation into rogue AI agents. Smith’s proposals would give a probe like that something firmer to enforce: a reporting duty that exists before the damage, not after a journalist finds it.
What to watch next
- A UK consultation. Any formal step from the government would show whether the minister’s remarks turn into policy
- Rival labs’ response. OpenAI, Google and Anthropic have not publicly backed binding pre-release sign-off
- Draft language. The proposal only matters once someone defines which models count as frontier and who holds the brake
Smith’s pitch is easy to remember, and that is partly the point. Nobody argues that milk should skip inspection. The harder question is who runs the inspection of a model that can write its own code, and Microsoft has now said, on the record, that it wants the answer to be a government.
Sources and further reading
- The Next Web: Milk is more regulated than AI, Microsoft’s Brad Smith tells the BBC
- TechCrunch: Microsoft president says UK is best placed to drive AI regulation
- TechCrunch: Microsoft’s Satya Nadella says AI models need an emergency brake
About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.
