U.K. AI minister Kanishka Narayan: 'harden defenses' on AI risk

The remarks came at the Labour Party's annual conference, where Narayan was asked directly about Coxon's warning by Fortune. Coxon's post—viewed 175 million times—claimed that people building AI earnestly believe it could kill us all by the end of the decade, and Anthropic scientist Evan Hubinger agreed, putting the risk above 10%. Bill Gates added his own warning on NBC's Meet the Press, saying that in the wrong hands AI was powerful enough to cause a billion deaths. Those statements set the backdrop for Narayan's response: rather than dismissing the fears, he said that when people in labs say they are very worried, you have to take that seriously.

Narayan's position sits between two poles. U.S. President Donald Trump has called talk of AI risks a hoax and warned that slowing U.S. frontier-lab research would hand China an advantage, while the White House is reported to have urged G20 members toward a "hands off" approach on global rules. Meanwhile, Democrats and Republicans have called for new legislation aimed at regulating labs, and U.K. Prime Minister Andy Burnham told the UN General Assembly that frontier labs were a "multiplier of risk," with the U.K. able to act as an honest broker. Narayan also listed his own priorities: cybersecurity risk, biochemical risk, and instances where models go beyond what humans have asked, while noting the "logical loops" needed to make it an existential question are significant and severe.

For U.K. officials, the bet is that the country's three years of testing work gives it standing to shape principles through next year's G20 presidency—though that hinges on whether other governments, particularly the U.S., engage rather than stay hands off. Narayan also stressed that organizations, individuals, government and the private sector have agency over whether AI augments or automates human labor, and that there are ways of adopting technology that mean more jobs and more productivity.