London Warns of Extinction-Level AI Risk While Washington Plays Chicken With China

(SeaPRwire) –

By: Marcus Sinclair

A former Anthropic employee posted a single warning on social media that has now accumulated 175 million views. Jacob Coxon wrote that the people building frontier artificial intelligence earnestly believe it could kill everyone by the end of this decade. The British AI minister, Kanishka Narayan, did not dismiss it. He told delegates at the Labour Party conference that when people inside labs say they are very worried, you have to take that seriously. Then he issued the policy prescription: harden and build your defenses.

That single word harden is doing enormous ideological work. It shifts the conversation from growth management to protective mobilization. It borrows language from the Cold War and applies it to a technology that does not yet have a fixed physical form. Narayan said testing is good but clearly insufficient. He said the risks have only grown. He positioned the U.K. to convene G20 nations next year under its presidency as the architecture of principle for global governance. These are not casual remarks from a mid-level bureaucrat. Narayan sits in the Prime Minister Cabinet.

Prime Minister Andy Burnham made an even sharper claim at the United Nations General Assembly last week. He called AI frontier laboratories a multiplier of risk. He said the U.K. could play the role of honest broker between competing national interests. That is a diplomatic ambition wrapped inside a security alarm. It assumes the world still believes multilateral institutions can steer runaway technology. It also assumes America will follow British leadership rather than forge its own path. Both assumptions are worth interrogating.

Narayan laid out his risk portfolio with unusual candor for a sitting minister. He cited a secular uplift in cybersecurity threats facing every organization and individual. He raised concerns about biochemical proliferation driven by tool democratization. He acknowledged models going beyond what humans asked them to do. He then immediately qualified each of these concerns by saying the logical loops required to make an existential threat real are quite significant and severe. That is not the language of someone who believes collapse is imminent. That is the language of someone who believes the threat landscape has expanded beyond any existing regulatory capacity.

The gap between London and Washington is now structural. Trump called AI risk talk a hoax. He warned that controlling the pace of American frontier lab research would hand China a decisive advantage. He said the U.S. leads China by a pretty good period and that losing AI means being put in a very bad position. The White House reportedly warned G20 members to take a hands off approach to new global regulations. This is not policy confusion. It is a deliberate strategy. American competitiveness in AI is being traded against American security architecture for AI.

Evan Hubinger, a scientist at Anthropic, backed Coxon by placing the existential risk above 10 percent. Bill Gates told NBC Meet the Press that AI in the wrong hands is certainly powerful enough to drive events causing a billion deaths. These are not fringe voices. They are people inside the labs watching the frontier expand. Narayan said exactly that phrase when he responded: you have to take that seriously.

The economic argument cut both ways. Narayan told a Resolution Foundation meeting that organizations and individuals have agency over whether AI augments labor or automates it. He said there are ways of adopting the technology that produce more jobs and more productivity. He admitted the U.K. had underplayed the role of innovation and technology IP in driving economic growth. This is a standard pro-innovation framing. It is also a standard way of refusing to confront the distributional consequences of automation at speed.

The geopolitical endgame is becoming legible. The U.K. wants to be the rule-maker. America wants to be the race winner. Both positions assume the other side will accommodate their preferences. Neither position has a mechanism for enforcing compliance on laboratories that answer to capital markets and shareholder returns rather than national governments. Narayan said the U.K. has been doing testing for three years and is very uniquely placed compared to any other country. That confidence assumes the rules Britain writes will be followed by companies headquartered in jurisdictions that explicitly rejected them.

The frontier lab industry operates across borders. Anthropic has American investors and American corporate structure. Its scientists publish risk assessments that cross Pacific and Atlantic audiences simultaneously. A British minister can warn about extinction risk and propose G20 architecture. An American president can call those warnings a hoax and instruct allies to keep their hands off. The technology does not pause for either position.

Narayan’s portfolio framing of risk is honest about complexity. It admits no single threat dominates. It accepts that existential scenarios require jumping through significant logical hoops. It refuses to ignore the warnings from inside the labs. The U.K. approach amounts to building defenses around a moving target while hoping other countries will help hold it steady. The American approach amounts to accelerating past the target and assuming speed is security. Both strategies treat governance as secondary to capability. That is the real gridlock. It is not a disagreement about facts. It is a disagreement about whether the race itself should be governed at all.

Author bio: Marcus Sinclair, a Senior Fellow at a prominent European geopolitical and security think tank specializing in technology governance and transatlantic strategic relations.