For much of the past few years, AI has been a financial story told through infrastructure booms, capex cycles and labour market disruption. This week was a reminder that behind all the projected productivity stats and Goldman Sachs white papers, it has always been a story about power.
Before the US and Israel launched their war on Iran, Anthropic — the safety-focused AI lab behind Claude — clashed with the US Department of Defence and Secretary Pete Hegseth over the terms under which the military could use its models.
At the centre of the dispute was Anthropic’s refusal to allow Claude to be deployed for fully autonomous weapons or the mass domestic surveillance of Americans.
That standoff ended badly. The Pentagon moved to designate Anthropic a supply chain risk and signed a deal with rival OpenAI instead, under terms that remain, let’s say, creatively vague. Anthropic was, in Donald Trump’s words, “fired like dogs”.
