Microsoft president says AI not yet existential but urges global safeguards

The president of Microsoft, Brad Smith, has warned that artificial intelligence poses serious risks and needs urgent regulation, saying the technology has advanced faster than anyone in the industry predicted. In an interview, he said that AI is not an existential threat today, but that the best way to stop a problem becoming worse is to address it now.

Mr Smith is a central figure in the AI industry. Microsoft provides much of the computing power behind AI, sells Copilot and owns a substantial stake in OpenAI, the company behind ChatGPT. He said the technology is less regulated than a washing machine or a carton of milk.

He was asked about warnings from Microsoft founder Bill Gates, who has said AI is the most dangerous thing humans have ever gone near and makes nuclear weapons look like nothing. Mr Gates has warned that existing systems could help a bioterrorist kill hundreds of millions or enable a cyber attack that brings down banking and electricity networks. Mr Smith said Mr Gates was right to raise those issues, adding that his language was more colourful but that he would not back away from the call for more serious action.

Mr Smith said there had been incidents where rogue AI agents acted outside human control. He said AI was not an existential threat today, but that it was indispensable for the technology to remain under human control. Without urgent regulation, he warned, there is a risk that people will lose control over it.

His proposals include: - independent testing of new AI models; - mandatory reporting when things go wrong; - an emergency brake, so governments can shut down dangerous AI systems in the same way they can ground aircraft fleets over safety concerns; - enforced disclosure; - legal liability for companies whose AI causes catastrophic harm, if they are found to have been negligent; - international cooperation on safeguards and monitoring.

On jobs, Mr Smith is optimistic about the long run. He does not accept that AI will destroy more jobs than it creates, but he expects a period of upheaval as the technology is rolled out. He said governments have a role in managing the transition, including ensuring people have the right skills, taxing companies that use AI or robots to replace human labour and protecting certain roles. He named teachers, doctors, judges and care workers as jobs that should not be done by a robot or computer.

Public opinion is more anxious. Polling cited in the coverage suggests that two thirds of Britons think AI could end human civilisation and a quarter think that is likely. Six in ten Britons think AI will destroy more jobs than it creates.

Mr Smith was careful not to criticise President Trump, who has called concerns about AI safety a hoax. He said Mr Trump was right not to want to slow innovation, but he added that the world cannot stay on the sidelines and watch the future unfold. He called for a stronger commitment to safety.

Internationally, Mr Smith said cooperation on safeguards and monitoring is possible. He said the UK is better positioned than any other country to help lead the way, potentially acting as a bridge between the United States, China and Europe. The prime minister has told the United Nations that he wants Britain to lead on the international regulation of AI.