Trending 
AI company Anthropic may lose a $200 million Pentagon contract due to its policy prohibiting the use of its technology for mass surveillance of US citizens or in fully autonomous weapons systems.
A US official told AFP on Monday that Anthropic executives are scheduled to meet with Pentagon officials on Tuesday to discuss the matter, first reported by Axios.
Anthropic is among the world’s leading AI firms, alongside OpenAI, creator of ChatGPT, and Google, whose Gemini models compete in capability.
The three companies, along with Elon Musk’s xAI, were contracted last year to provide AI models for a range of military applications, described by the Pentagon as “frontier AI” projects.
While the Pentagon expects all four companies to make their models available for any applications permitted under US law, Anthropic has set explicit boundaries for certain uses, prompting concern from military officials.
Reports suggest the Pentagon may even consider barring Anthropic from future defence contracts.
“You can’t have an AI company sell AI to the Department of War and [then] don’t let it do Department of War things,” said Emil Michael, a senior adviser who acts as the US military’s de facto chief technology officer, speaking to DefenceScoop.
The term “Department of War” comes from a Trump administration executive order; formally renaming the Department of Defence would require Congressional approval.
Founded in 2021 by former OpenAI employees, Anthropic has maintained a focus on AI safety while competing in commercial and government markets.
The company declined to comment directly on its red lines but emphasised its commitment to supporting US national security and said it had “productive conversations, in good faith, with DoW on how to continue that work and get these complex issues right.”
According to Axios, talks between Anthropic and defence officials are nearing a breakdown, marking the latest tension between the company and the Trump administration, which has criticised Anthropic for adopting what it considers an overly cautious stance on AI.
