On a Sunday that might otherwise have passed quietly, President Donald Trump invited Dario Amodei, the chief executive of the artificial-intelligence laboratory Anthropic, to the White House for a meeting described by officials as an attempt to narrow deep disagreements over the future direction of AI development. The encounter was notable not only for its timing but for its context: Anthropic had recently been barred by the Pentagon from certain defence-related contracts, a decision that placed the San Francisco-based company in an uncomfortable position between government suspicion and Silicon Valley ambition. The Pentagon's ban on Anthropic was not explained in full detail publicly, yet it sent an unmistakable signal. Among the most advanced AI laboratories in the world, Anthropic occupies a singular space — founded explicitly around the principle of AI safety, and yet now caught in the crossfire of a broader struggle over who controls, and who is trusted to develop, the most consequential technology of the era.

Amodei, a former OpenAI researcher who co-founded Anthropic with his sister Daniela in 2021, has long argued that safety and capability need not be in conflict. Whether Washington currently shares that view is, at best, uncertain. The meeting reflects a wider pattern visible across two continents. In Brussels, European regulators are tightening rules on foundation models, while in Washington the instinct has been to treat frontier AI simultaneously as a strategic asset and a potential threat.