President Trump’s Department of War is ready to sever its relationship with the major AI developer, Anthropic. The administration is threatening to cut ties because Anthropic won’t agree to let the Pentagon use its advanced AI systems for mass domestic surveillance or for creating fully autonomous weapons, as reported by The New Republic.
This whole fight stems from the fact that the Department of War is apparently fed up with Anthropic’s refusal to allow unfettered use of its systems. The AI company actually signed a contract with the Pentagon last summer that was valued at up to $200 million. However, when they signed that deal, Anthropic was very clear about setting hard boundaries. They explicitly said their systems couldn’t be used for developing weapons, facilitating violence, or conducting surveillance.
For months now, the Pentagon has tried to negotiate, wanting the ability to use the tools for “all lawful purposes.” This includes intelligence gathering and weapons development, without having to argue over every single use case. The administration views this restriction as completely unacceptable.
It’s a tough spot for a company trying to balance innovation with ethics
A senior Trump administration official stressed that “Everything’s on the table,” when discussing the future of the partnership, including cutting ties altogether. This sounds like the administration is issuing a clear ultimatum: play ball or you’re out. If the relationship does end, the official noted that they would need to find an “orderly replacement” for Anthropic’s technology.
Chief Pentagon spokesman Sean Parnell also weighed in, suggesting the relationship was already under review. Parnell made it clear where the administration stands, stating, “Our nation requires that our partners be willing to help our warfighters win in any fight.”
Tensions really came to a head back in January, according to the official. An executive at Anthropic reportedly reached out to an executive at Palantir regarding a specific U.S. military raid. This raid was aimed at capturing Venezuelan President Nicolás Maduro. The Anthropic executive wanted to know if Palantir had used the Claude AI assistant as part of that operation, and spoiler alert, it had.
Since kinetic fire was used during that raid and people were shot, the senior official claimed that the inquiry implied Anthropic “might disapprove of their software being used.” This type of ethical oversight is obviously not what the Department of War wants when it’s trying to deploy technology in the field.
Anthropic, however, strongly denies that this specific conversation about the Maduro raid ever took place. A company spokesperson issued a statement saying they had “not discussed the use of Claude for specific operations with the Department of War.” They added they hadn’t discussed it with any industry partners either, outside of routine technical discussions.
The spokesperson clarified that Anthropic’s current conversations with the Department of War are focused entirely on their usage policy questions. Specifically, they are sticking to their hard limits regarding “fully autonomous weapons and mass domestic surveillance.”
Published: Feb 17, 2026 02:30 pm