Microsoft President Brad Smith is calling for global safety rules for artificial intelligence, including emergency shutdown mechanisms, mandatory incident reporting and independent reviews of some products before launch. He says AI currently faces fewer safety requirements than products such as washing machines and milk, raising questions about companies’ responsibilities when systems fail and users’ right to know when they are affected.
Smith calls for incident reporting and emergency shutdowns
In an interview with the BBC’s Big Boss Interview, Smith described AI as both a “tool” and a “weapon” and said regulation should begin now. He argued that companies should be able to pause or slow systems when they malfunction or cause serious problems. Such a mechanism is often called a “kill switch.”
Smith also proposed legally required incident reporting. Companies, he said, should notify affected people when an AI agent causes harm or a system is compromised. For some new AI products, he suggested that independent government assessors review them before they reach the market.
He pointed to existing product-safety rules as a potential model. “Today, washing machines are subject to more safety rules than AI,” Smith said. “Every carton of milk in a grocery store has to comply with safety regulations and be inspected.”
US and UK differ on how to regulate AI
Smith’s position contrasts with US President Donald Trump’s preference for industry self-regulation. Trump has argued that US companies should not be held back by rules while competing with China, and has repeatedly said that “whoever wins AI will own everything.” Smith said voluntary controls alone cannot go far enough, particularly when companies are unwilling to take responsibility.
Debate over the pace of AI development and the need for safeguards has intensified. Leaders at US AI companies Anthropic, OpenAI and Google DeepMind have backed proposals to slow development, while Trump has opposed doing so. Smith warned that rules are needed if AI agents move beyond human control or become powerful enough to threaten global cybersecurity.
He said the UK could help drive international discussions, citing its AI safety research capabilities. The country’s AI Safety Institute was established under former Prime Minister Rishi Sunak and is now called the AI Security Institute. Smith said people in the technology sector see the UK as a leader in relevant technical capabilities.
On whether Washington would accept rules led by the UK, Smith said US politicians view London differently from Brussels. In his view, that could make standards proposed by the UK more acceptable to the US. He did not say that the two countries had agreed on a specific regulatory framework.
Smith says White House company pledge is not enough
Late last month, Trump met leaders from OpenAI, Anthropic, Nvidia, SpaceX, Meta and Google at the White House. The companies signed a voluntary, ethics-based pledge to work with “independent auditors” to check whether AI systems perform as intended and do not access or intrude on technical infrastructure in unintended ways.
Smith called the pledge a “good and necessary step” and said Microsoft would be willing to sign it. But he argued that voluntary commitments are no longer sufficient. Companies that already take compliance seriously are likely to adopt safeguards; the harder problem is how to address firms that do not.
The policy debate therefore extends beyond how AI models are evaluated. It also concerns who conducts those assessments, what companies must report and disclose after an incident, and whether rules can cover firms that do not regulate themselves. Smith’s call for global AI safety rules remains a proposal; specific arrangements have yet to be settled.