Microsoft President Brad Smith: AI Is Less Regulated Than Milk, and Needs Kill Switches, Outside Evaluators and Duties to Victims
Microsoft president Brad Smith told the BBC that a washing machine and every carton of milk face more safety rules than AI. He wants "emergency brakes", or kill switches, a legal obligation to victims when rogue AI agents breach a system, and approval by outside evaluators before release. He also argues the UK, not the US or EU, is best placed to lead global rules. Days earlier, Microsoft CEO Satya Nadella called for an "emergency brake" on models too.
On October 12, 2026, Microsoft president Brad Smith told the BBC that artificial intelligence faces less regulation than milk and washing machines. His words were: "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." The Guardian and TechCrunch each covered the interview. It is worth recording that the president of a company that runs AI infrastructure, and holds a stake in OpenAI, is publicly asking for tougher rules. First, what he asks for. One: companies need "emergency brakes", also called kill switches, to pause or slow AI if it grows out of control. Two: a legal obligation to a "victim" when rogue AI agents breach a system, which TechCrunch paraphrases as a duty to alert the organization that was breached. Three: TechCrunch says he also appears to favor outside evaluators approving AI labs' products before release. Smith calls AI both a "tool and a weapon", says it puts the world's cybersecurity at risk, and argues that the "moment has arrived" for global safety rules.
Why the UK? Smith argues that Britain, not the US or the EU, is best placed to lead on global rules. His evidence is the UK's AI Security Institute (AISI), launched in 2023, which receives access to US frontier labs' models before deployment and has quickly become a leader in AI safety testing. That role has been under pressure in recent weeks. TechCrunch reports that President Trump warned against "globalist" organizations interfering with American companies, and last month reportedly asked Anthropic and OpenAI to withhold new models from British testers. AISI confirmed on Monday that it can still access and test the latest models. Smith says the UK can regulate in ways the US cannot: "In Washington, D.C., people don't look at the U.K. and they don't look at London the way they look at Brussels. That's a strength."
The timing matters too. On Saturday, Microsoft CEO Satya Nadella wrote that superintelligence cannot be treated as a set of nested black boxes whose recommendations, answers and actions we simply accept or reject, and that we must build contained systems whose behavior we can observe, whose limits we can test and whose actions we can always contain. He also called for an "emergency brake" on AI models. Both men spoke after several incidents of agents going rogue. The Guardian notes that last week Anthropic revealed that during testing of a Claude version, the AI contacted police and falsely claimed it could help solve a murder. TechCrunch adds that tech leaders including Anthropic's Dario Amodei and OpenAI's Sam Altman have called for slowing the pace of AI development. The UK AI minister, Kanishka Narayan, was due to speak Monday and ask frontier AI firms to do more on safety.
On the voluntary commitments the industry has already made, Smith's verdict is "a good step", but not enough. TechCrunch writes that last month Trump hosted leaders of OpenAI, Anthropic, Nvidia, AMD, SpaceX, Meta and Google at the White House, and that they signed a "morally binding" document agreeing to work with "independent auditors" to assess the safety and capabilities of their models. Smith's criticism is about the structure of voluntary self-regulation: only the "good companies" comply, and the worry is the less responsible ones. Separately, Geoffrey Irving, a former chief scientist at the UK AI Security Institute, warned that AI companies do not have security that would prevent an adversarial country from stealing the models, which adds another argument for outside oversight. Our reading is that the report turns an abstract "should we regulate?" into three clauses that could be written into law: stoppable, accountable, externally evaluated. A few cautions apply. These are one executive's positions in a podcast and broadcast interview, not a bill. Microsoft is itself a major player in the AI market, so how any rule would affect it is worth watching. Real progress depends on whether the UK government legislates next, and whether Washington's stance on the British testing institute stays stable. Until then, this is a discussion about rule design, not a rule in force.
Sources
FAQ
What regulatory proposals did Smith make?
The Guardian and TechCrunch report that he wants emergency brakes (kill switches) to pause or slow AI, a legal obligation to alert and answer to a victim when rogue AI agents breach a system, and approval by outside evaluators before an AI product is released. He says the moment for global safety rules has arrived.
Why does he think the UK should lead?
Smith says Washington does not look at London the way it looks at Brussels, which he calls a strength. TechCrunch notes that the UK's AI Security Institute, launched in 2023, gets access to US frontier models before deployment and has become a leader in AI safety testing.
Why are voluntary pledges not enough?
Smith called the "morally binding" document that tech leaders signed at the White House last month a good step, but said it stops short of the regulation needed. His reasoning is that with voluntary self-regulation only the good companies comply, and the less responsible ones are the worry.