Microsoft President Brad Smith: AI Is Less Tightly Controlled Than Milk

TVS News AI, editorially supervised, presents Microsoft President Brad Smith: AI Is Less Tightly Controlled Than Milk.

In a BBC podcast "Big Boss Interview," Microsoft President Brad Smith drew a comparison that is hard to ignore: "Every milk in the supermarket is monitored more strictly than AI." The technology, he said, is both a tool and a weapon, and urgently needs to be regulated — ideally immediately.

Smith is now the third top manager at the software company to raise the alarm loudly within just a few days. On October 10, 2026, CEO Satya Nadella used a social media post to call for stricter controls on AI systems and a kill switch. Microsoft co-founder Bill Gates had put it even more starkly to the New York Times: AI is the most dangerous thing humanity has ever dealt with — nuclear weapons, by comparison, are small fry. The danger threshold, Gates said, had already been crossed in early 2026.

Every milk in the supermarket is regulated more strictly.

What Smith specifically demands of policymakers and industry

Smith doesn't stop at the comparison. He demands the same emergency brakes Nadella brought into play — meaning shutdown mechanisms for AI systems in a worst-case scenario. There should also be a legal obligation for companies to report possible incidents and inform those affected. Independent government inspectors should review new AI technologies even before release, the way it is standard practice with other products. Speaking to the British broadcaster ITV, Smith added: In the event of an AI-triggered catastrophe, the company responsible should be held liable — provided it acted negligently. The central challenge, he said, is to make sure humans remain in control.

The contradiction with the US government

The fact that the very top brass of an AI company is calling so loudly for regulation has an awkward flavor: Microsoft makes billions from AI. At the same time, US President Donald Trump, at a meeting with leading US providers, said the tech companies should more or less regulate themselves — otherwise they would lose the race against China. Smith's demand thus also reads as an open counter-position to his own government.

What Microsoft itself already delivers

According to Cnet, the company already has a product on the market based on these considerations: so-called Execution Containers. The enterprise security feature restricts AI models' access to certain files and programs and comes with an emergency switch that immediately stops the models when needed. Smith said in the BBC podcast: "We should build on this knowledge and think about how we can apply it to AI."

What this means for you

When three of the most powerful minds at an AI company within just a few days almost unanimously call for stricter rules, that is no marketing slogan. It is an admission that, in the view of the developers themselves, the current self-regulation is not enough. So you are hearing from the very people who know the technology best that they want more rules — not fewer.