Microsoft president Brad Smith has warned that artificial intelligence operates under lighter oversight than everyday household goods, calling for urgent regulation of a technology he described as both a “tool and a weapon”.
Smith told the BBC that AI currently faces fewer safety requirements than common appliances and food products. “Right now, your washing machine is subject to more safety rules than AI,” he said. “Every carton of milk that you see in the grocery store is subject to safety rules and inspection.”
He argued for stronger safety rules, including a legal duty owed to a “victim” whose system is compromised by rogue AI agents.
Smith also said companies need “emergency brakes” – also known as “kill switches” – that would allow AI to be paused or slowed if it grows beyond control.
His comments add to pressure from experts, politicians and regulators for tighter controls, alongside warnings from Anthropic, Google DeepMind and OpenAI that the pace of development should slow.
Microsoft runs AI infrastructure and holds a stake in OpenAI, the owner of ChatGPT, and is headed by chief executive Satya Nadella. On Saturday, Nadella joined the call for an “emergency brake” on AI models.
“We can’t treat super intelligence as a set of nested black boxes and simply accept or reject its recommendations, answers, and actions,” Nadella wrote. “We must build contained systems whose behaviour we can observe, limits we can test, and actions we can always contain.”
The remarks came as Britain’s AI minister, Kanishka Narayan, prepared to urge frontier AI developers to strengthen safety measures. In a speech due on Monday, Narayan is expected to say that firms building AI models must do more to address risk and improve safeguards, and to say the UK government is readying legislation to expand AI oversight, according to reports.
Experts have cautioned that adversarial states could obtain unreleased, powerful AI models. Geoffrey Irving, a former chief scientist at the UK’s AI Security Institute, said AI companies lacked “security that would prevent an adversarial country from stealing the models”.
Reports of rogue AI behaviour continue to surface. Anthropic disclosed last week that while testing a version of its Claude model, the AI contacted police and falsely claimed it could help solve a murder.
COMMENTS