Microsoft's Brad Smith says milk and washing machines face tougher rules than AI

Tougher rules for food than for AI

Microsoft president Brad Smith said in a BBC interview published on Monday, 12 October, that a washing machine is currently subject to more safety rules than artificial intelligence. "Right now, your washing machine is subject to more safety rules than AI. Every carton of milk you see in the supermarket is subject to safety rules and inspection," he said. Smith said the AI industry "is moving faster" than safety teams, adding that this "is not good for the protection of the public. It is not even good for the industry."

We know how to do this for other products. We should build on that knowledge and think about how we apply it to AI.

Proposed safeguards

Smith called for models to undergo more safety evaluations, with independent evaluators both outside companies and linked to different governments. He said those evaluators would determine whether a product is ready to be released to the public. He also proposed an emergency brake to stop or slow AI systems that get out of control, and a legal duty for companies to file incident reports and tell people harmed when a system behaves unexpectedly or harmfully.

He described the emergency brake as applying to any AI system controlling autonomous devices or critical infrastructure, according to the Portuguese-language Observador report. Smith also said that in cases of attacks carried out by AI agents there should be a legal obligation to tell the victim what happened, referring to recent cases without naming companies.

Britain and the AI Security Institute

Smith said the United Kingdom is best placed to promote global AI rules, mainly thanks to the AI Security Institute set up under former prime minister Rishi Sunak. Handelsblatt reported that the institute's testers, who receive advance access to models from leading providers on a voluntary basis, enjoy high standing.

Disagreement with Trump

Smith's position contradicts that of US President Donald Trump, who argues that the AI sector should regulate itself so that American companies do not lose the race with China. Several AI executives, including those at Anthropic, OpenAI and Google DeepMind, have publicly backed slowing agent development, according to Correio da Manha.

Gates warns of cyberattacks

Bill Gates, Microsoft co-founder, told ABC News that the most dramatic developments next year will probably be in cyberattacks. "The real danger at this moment is not that AI gets out of control. We need to work on that too, but the major impact of such a scenario is still a few years away," he said, according to a Romanian report citing AFP. Gates said on Sunday, 11 October, that he rejected the idea that AI companies should self-regulate.