Microsoft president Smith has called for urgent legislative action to regulate artificial intelligence and require developers to provide mechanisms that stop or slow their systems when they become uncontrollable, alongside incident reporting and independent assessments of some products before they are launched on the market.
Smith calls for urgent AI rules
He described the technology as “a tool and a weapon at the same time”, stressing that work to regulate it could not be delayed. Smith told the BBC’s Big Boss Interview podcast that conventional consumer products are currently subject to stricter controls than artificial intelligence.
“At present, a domestic washing machine is subject to more safety rules than artificial intelligence, and every bottle of milk sold in shops is subject to inspection procedures and safety standards,” he said. The Microsoft president stressed the need to begin establishing rules immediately, saying: “We need to start today”.
His call comes amid growing demands in recent years for tougher laws to limit the risks posed by artificial intelligence, following warnings from leading experts in the sector about the potential threats the technology could pose to humanity.
Emergency brakes and mandatory system reports
Smith said developers needed to equip their systems with what he described as “emergency brakes” — mechanisms known in the sector as “kill switches” — enabling a system to be stopped or slowed if it began behaving in a way that exceeded human control or the limits set for it.
He also called for legal rules requiring companies to prepare incident reports and notify affected people or organisations when systems behave unexpectedly. The cases he referred to included damage or breaches caused by “rogue AI agents” that escape control. He also called for the types of products and systems that should not reach the market without approval from an independent assessment body to be identified.
He said reviews of new technologies before their launch could be assigned to government or independent bodies, adding: “We know how to apply these procedures to other products, and we should build on that experience and think about how to apply them to artificial intelligence”.
Britain seeks to lead international standards
Smith said the United Kingdom had an advantage that enabled it to lead international efforts to establish regulatory standards for artificial intelligence. In this context, he cited the AI Security Institute, which was launched under the government of former British prime minister Rishi Sunak.
He said people working in the technology sector viewed Britain as a leading country in the technical capabilities linked to this issue, adding: “In the technology sector, if you ask most people in it which country is leading in this area in terms of technical capabilities, they will simply say: the United Kingdom. It is the best”.
Dispute between binding legislation and self-regulation
Smith’s call for binding legislation differs from the position of US president Donald Trump, who is pushing for the sector to regulate itself, fearing that restrictions could cause US companies to lose their advantage over Chinese competitors. Trump is also calling for investment and development in artificial intelligence to be accelerated, and has repeated that “whoever wins AI wins everything”.
By contrast, leaders of major US companies working in artificial intelligence, including Anthropic, OpenAI and Google DeepMind, have backed slowing the pace of development of some advanced systems, while Trump continues to push for faster investment and innovation in the sector.
The Microsoft president does not believe the US administration would completely reject regulatory initiatives led by Britain. He said Washington might respond more positively to British proposals than to rules issued by the European Union, saying: “In Washington, they do not view London in the same way as they view Brussels, and that is a strength”.
He added that the need for action was growing as AI agents emerged that could escape human control, and as highly powerful models were developed that could threaten global cybersecurity. He said: “When you see AI agents that have escaped human control, and when you look at extremely powerful models that could threaten global cybersecurity, I say the time to act has arrived”.
Smith called for the debate over regulation to be conducted calmly and firmly, bringing different parties together around shared rules. He added: “We need to talk calmly but firmly, and bring the different parties together around this issue. I hope that this is the role Microsoft can play, and that the United Kingdom can play as well”.
At the end of last month, Trump hosted leaders of several major US technology companies at the White House, including OpenAI, Anthropic, Nvidia, SpaceX, Meta and Google. The participants signed a document described as “ethically binding”, under which the companies pledged to work with independent auditors to assess AI systems and ensure that they operated in line with their specified objectives.
The document also included a commitment to prevent AI systems from breaching or accessing technological systems in unintended ways. Smith praised the initiative, stressing that Microsoft was pleased to join it, but described it as a beginning that could not replace binding legislation, saying: “It is a good and necessary step, but it is no longer sufficient on its own”.
He said the main problem with self-regulation was that responsible companies would comply with the required rules and standards, while concerns remained about companies that were less committed to acting responsibly.
From this standpoint, Smith believes that voluntary commitments need a legal framework requiring reporting and independent assessment, and providing practical means of stopping systems when they display behaviour that is out of control.
