Anthropic and the Pentagon dispute highlights how AI is used in warfare

In short: A public dispute between Anthropic and the US Defense Department is drawing attention to how AI is already built into modern military operations.

What's going on

AI, meaning software that can spot patterns and make suggestions, has been part of US military work for years. One major example is Project Maven, which started in 2017 and used AI to scan drone video faster (like having a tireless assistant watching many screens at once). Google worked on it at first, but employee protests pushed Google to step away in 2018. Other companies later continued the work, and it evolved into a system called Maven Smart System.

The current flashpoint involves Anthropic, the company behind the Claude chatbot. The Verge reports that in January 2026, Defense Secretary Pete Hegseth pushed to rewrite AI contract terms so the Pentagon could use the tools for “any lawful use.” Anthropic objected and has said it wants two “red lines,” no domestic mass surveillance and no weapons that can find, track, and kill targets with zero human involvement.

Even without fully autonomous weapons, experts quoted by The Verge say AI can speed up the “kill chain,” meaning the steps from spotting something to acting on it, so much that humans may only be rubber-stamping decisions. US policy on autonomous weapons exists, but key terms remain fuzzy, and international talks have moved slowly.

What to watch

Anthropic’s limits may not set the standard for the rest of the industry. The Pentagon has signed deals with several other major tech firms to deploy AI on classified networks, which could reduce Anthropic’s leverage and increase pressure to move faster.

Source: The Verge AI