The Pentagon used Anthropic’s Claude AI model during the U.S. military raid that captured Venezuelan leader Nicolás Maduro, according to the Wall Street Journal and Axios. The revelation has exposed deep tensions between Anthropic’s safety-first ethos and the Defense Department’s push for unrestricted AI access in military operations.
The U.S. military’s use of Claude during an active classified operation marks a watershed moment in the intersection of artificial intelligence and modern warfare. Two sources with knowledge of the situation confirmed to Axios that Claude was deployed during the operation itself — not just in preparatory intelligence work.
What Happened
U.S. forces conducted a raid in Caracas, Venezuela to capture Nicolás Maduro, an operation that involved bombing across the capital and resulted in 83 deaths according to Venezuela’s defense ministry. Claude was reportedly accessed through Anthropic’s partnership with Palantir Technologies, a major defense and intelligence contractor.
The precise role Claude played remains unclear. The military has previously used Claude to analyze satellite imagery and process intelligence. Sources indicated it was used during the active operation, suggesting real-time analytical support rather than background research.
The Anthropic Dilemma
Anthropic’s terms of use explicitly prohibit deploying Claude for violent purposes, weapons development, or surveillance. The company has positioned itself as the industry’s safety-first leader, with CEO Dario Amodei repeatedly calling for AI regulation and expressing wariness about autonomous lethal operations.
“Anthropic asked whether their software was used for the raid to capture Maduro, which caused real concerns across the Department of War indicating that they might not approve if it was,” an official told Axios. The company declined to confirm or deny Claude’s involvement but said any use must comply with its usage policies.
Pentagon Pushes Back
The Defense Department wants AI companies to allow unrestricted military use so long as operations comply with the law — a far broader standard than Anthropic’s internal policies. Secretary of War Pete Hegseth said in January the Pentagon wouldn’t “employ AI models that won’t allow you to fight wars.”
The Pentagon has been diversifying its AI partnerships. It announced a deal with Elon Musk’s xAI in January and already uses custom versions of Google’s Gemini and OpenAI systems for research support.
Broader Implications
The incident underscores a growing rift between Silicon Valley’s AI safety commitments and Washington’s operational demands. Israel’s military has already deployed AI-powered autonomous drones in Gaza and used AI extensively for target selection. The U.S. has used AI targeting for strikes in Iraq and Syria.
Critics warn that AI in weapons systems creates targeting errors with lethal consequences. But the defense establishment views AI reluctance as a competitive liability — particularly as adversaries accelerate their own military AI programs.
Anthropic is currently negotiating terms of use with the Pentagon, seeking guarantees against mass surveillance of Americans and fully autonomous weapons. The outcome could set the template for how every major AI company engages with the defense sector.
Sources: Axios, The Guardian
