UK AI Minister: Nations Must Harden Defenses Against Frontier AI Risk

The United Kingdom's minister for artificial intelligence has warned that nations must "harden and build" their defenses against AI risk, after a former Anthropic employee claimed a human extinction-level event could occur within the decade.

Kanishka Narayan, who sits in the Prime Minister's Cabinet, was responding to a claim by Jacob Coxon, a former Anthropic worker. Coxon wrote in a post on X, seen 175 million times, that "the people building AI earnestly believe that it could kill us all by the end of the decade."

Narayan took the warning seriously. "The first thing to say is that when people in labs tell you that they're very worried about this, you have to take that seriously," he said at the annual conference of the Labour Party, the governing party in the U.K. "You have to harden and build your defenses. The risks have only grown, so testing is good but clearly insufficient."

The warnings are not confined to one departed employee. Evan Hubinger, a scientist at Anthropic, agreed with Coxon and put the risk at greater than 10%. Over the weekend, Microsoft founder Bill Gates told NBC's Meet the Press that in the wrong hands, AI was "certainly powerful enough to drive events that, you know, cause a billion deaths."

A bid for global leadership

Narayan said the U.K. intends to use its upcoming G20 presidency to shape international AI rules. "We want to lead the world in being able to convene [countries] from the G20 next year as part of our presidency," he said. "Of course, we want to be fundamental drivers of the architecture of principles that are going to hopefully come out of that."

Prime Minister Andy Burnham told the United Nations General Assembly last week that AI frontier labs were a "multiplier of risk" and that the U.K. could act as an honest broker between competing national interests.

The regulatory push faces resistance from Washington. U.S. President Donald Trump, who has called talk of AI risks a hoax, warned that controlling the pace of American frontier lab research would hand China a significant advantage. "I have concerns that if we don't win AI, we're going to be put in a very bad position. We are leading China right now by a pretty good period," Trump said. The White House is reported to have warned G20 members to take a "hands off" approach to new global regulations.

In the U.S., Democrats and Republicans have both called for new legislation aimed at regulating the labs.

A portfolio of risks

Narayan set out his own risk assessment, one that weighs near-term threats more heavily than extinction scenarios.

"The candid sense I have is that I'm much more focused on risks today in terms of a secular uplift of cybersecurity risks that each of us in our organizations and lives," he said. "[I am] concerned about the proliferation of these tools in a way that drives up biochemical risk as well, and yes, I am concerned about instances where models go beyond what humans have asked them to do as well."

On the extinction question itself, he was blunt. "The sort of logical loops you have to jump through to make that an existential question are quite significant and severe, and so that's an indication of how I think about the portfolio of risks here."

On testing, Narayan claimed a global lead for Britain. "You've got to do the best testing in the world to be able to understand the risks and mitigate them. We're doing that. We've been doing that for three years. We're very uniquely placed compared to any other country to be able to do that."

The upside case

Narayan also pointed to gains from AI, including job creation, job searching and productivity. He argued the U.K. had underplayed the role of innovation and technology IP in driving economic growth.

Speaking at a meeting organized by the Resolution Foundation, a think tank, he framed the outcome as a matter of choice rather than inevitability. "We, as organizations, as individuals, as government, as the private sector, we actually have agency in whether this ends up being a technology that augments human labor, or is a technology that automates human labor," he said. "There are ways of doing technological adoption that means more jobs and more productivity."

With the U.K. aiming to anchor G20 AI principles next year against American resistance, the gap between British ambitions and Washington's "hands off" stance will define how far those defenses actually get built.