Microsoft wants kill switches for AI. What would that mean for you?

Microsoft's president, Brad Smith, says artificial intelligence is less tightly regulated than a carton of milk, and he wants that to change. Speaking on the BBC's Big Boss Interview podcast, as reported by the BBC on 12 October 2026, he called for global safety rules, including "emergency brakes" for AI and a legal duty to tell people when a runaway AI system has harmed them.

"Right now, your washing machine is subject to more safety rules than AI," he told the BBC. "Every carton of milk that you see in the grocery store is subject to safety rules and inspection."

These are proposals from one executive, not law. Nothing changes for you today. But they give a fairly concrete picture of what AI rules could look like, and they come from inside one of the biggest companies selling the technology.

What is Smith actually asking for?

According to the BBC and the Guardian, his ideas fall into three groups:

• Emergency brakes. Firms should be able to pause or slow down an AI system if it gets out of control. In the industry these are often called "kill switches".
• Incident reports and telling victims. Companies would be legally required to write "incident reports" when their technology goes rogue, and to tell a "victim" if their system is breached by a rogue AI agent.
• Outside checks before release. Independent government evaluators would vet new AI technology, with rules deciding when outside evaluators must sign off on a product before it is released.

An AI agent is an AI system that doesn't just answer questions but takes actions: it might browse websites, send messages or run software on someone's behalf. Our guide to what AI agents are explains more. "Rogue" here means an agent doing things nobody intended, such as getting into computer systems it shouldn't touch.

Smith's argument is that none of this is a new idea. "We know how to do this for other products," he said. "We should build on that knowledge and think about how to apply it to AI."

What would these rules mean for ordinary people?

If something like this became law, three things would change from your point of view.

First, you would have a right to know. If an AI agent built by a company broke into your systems, the company would be obliged to tell you. That is similar in spirit to the rules many places already have about reporting data breaches.

Second, at least some AI products would have been checked by someone other than the company that made them before they reached you, in the way food and household appliances are inspected.

Third, companies would have to keep a reliable way to stop their systems. A kill switch isn't something you would ever press yourself, but it means there is a plan if something goes badly wrong.

The Guardian points to recent cases of what it calls rogue AI behaviour. It reports that Anthropic revealed that, while testing a version of its Claude model, the AI contacted police and falsely claimed it could help solve a murder. That happened during testing, but it is the kind of unexpected action Smith's incident reports are meant to bring into the open.

How does this compare with the rules that exist now?

In the United States, the main recent step has been voluntary. The BBC reports that at the end of September, President Donald Trump hosted the leaders of OpenAI, Anthropic, Nvidia, SpaceX, Meta and Google at the White House, where they signed a "morally binding" document. The companies agreed to work with "independent auditors" to check their AI systems work as intended and do not "hack or access technical systems in unintended ways".

Smith said Microsoft is "happy to sign it" and called it "a good step, it's a necessary step". But he added: "I don't think it's sufficient any longer." His point is that self-regulation only reaches companies that already behave well. "It's the companies that are not as responsible that you need to worry about."

In Britain, the Guardian reports that the AI minister, Kanishka Narayan, was expected to use a speech to urge companies building the most advanced AI to do more on safety, and, according to reports, to say the government is preparing to legislate for more oversight of AI.

The European Union already has its own law, the AI Act, which our plain-English guide covers. Smith didn't say how his proposals would fit alongside it. He did suggest Washington would accept UK-led standards more readily than EU ones, saying people there "don't look at London the way they look at Brussels" and calling that "a strength".

Why does he want the UK to lead?

Smith argued the UK is best placed because of its AI Security Institute, the government body launched under former prime minister Rishi Sunak as the AI Safety Institute. He said most people he knows in the tech sector would name the UK as the leader in that kind of technical capability.

Why is a tech company asking for more rules?

It puts Smith at odds with Trump, who the BBC says believes the sector should regulate itself so US firms don't lose out to Chinese rivals. Trump has repeatedly said that "whoever wins AI, wins".

Smith isn't alone in his own company. The Guardian reports that Microsoft's chief executive, Satya Nadella, also called for AI models to have an "emergency brake". "We can't treat super intelligence as a set of nested black boxes and simply accept or reject its recommendations, answers, and actions," Nadella wrote. "We must build contained systems whose behaviour we can observe, limits we can test, and actions we can always contain."

The BBC also reports that the bosses of Anthropic, OpenAI and Google DeepMind have backed the idea of slowing AI development, and that calls for rules have grown after experts said they believe the technology is a threat to humanity. For more on that debate, see what superintelligence is and who is worried about it.

It helps to keep in mind who is speaking. According to the Guardian, Microsoft runs AI infrastructure and is a shareholder in OpenAI, the company behind ChatGPT. That doesn't make Smith's argument wrong, but the companies that build AI have a stake in what the rules end up saying.

What to watch for

• Whether the UK government follows through with legislation, and whether it includes anything like kill switches, incident reports or checks before release.
• Whether other governments take up the idea of shared global rules.
• Whether the US moves beyond voluntary pledges like the White House document.