US Defence Secretary Pete Hegseth has summoned Dario Amodei, the CEO of AI firm Anthropic, to the Pentagon for what officials describe as a high-stakes and potentially “tense confrontation” over military access to the company’s Claude Artificial Intelligence (AI) system, according to media reports.
The Tuesday meeting comes amid escalating pressure from the Defence Department on leading AI developers to make their most advanced models available inside classified US military networks with fewer restrictions.
An Axios report, citing informed sources, said that the US wants broader operational use of Claude — currently the only frontier AI model deployed in certain sensitive defence and intelligence environments — while Anthropic has sought to retain safety guardrails governing how its systems are used.
“Anthropic knows this is not a get-to-know-you meeting,” a senior defence official told Axios. “This is not a friendly meeting.”
The official said the Pentagon wants clarity on whether the company is prepared to support expanded military deployment on terms set by the government.
The dispute reflects a widening rift between US national-security agencies and Silicon Valley over how far powerful AI tools should be integrated into war-fighting, surveillance and intelligence workflows.
Earlier this month, Reuters reported that the Pentagon had been urging major AI firms, including OpenAI and Anthropic, to enable their models on classified systems without many of the usage constraints imposed on commercial customers.
Claude’s presence on classified networks has made Anthropic a pivotal partner in the Pentagon’s AI ambitions.
Officials have said the model’s performance on secure data and its ability to handle complex analytical tasks have made it the most capable system currently available in those environments.
Tuesday’s Pentagon meeting comes over a month after reports surfaced that Anthropic’s Claude AI was deployed during the January 3 US military operation in Venezuela that led to the capture of former President Nicolas Maduro.
Exactly how Claude was used remains classified. Neither operational details nor the AI’s specific role have been disclosed.
Anthropic has, however, refused to confirm the reports, stressing that its technology is bound by strict usage policies that prohibit violence-related applications, weapons development or surveillance.
