U.S. Department of Defense and Anthropic Fail to Reach Agreement on Military Use of Claude AI
The U.S. Department of Defense and AI technology developer Anthropic have failed to reach an agreement on the use of the Claude neural network for military purposes. According to Axios citing sources, the disagreements concern the principles of applying artificial intelligence in the defense sector.
The Pentagon intends to actively utilize neural networks for combat tasks—from operational planning to guiding precision weapons. The military department insists on lifting the restrictions that companies typically include in their user agreements. In turn, Anthropic firmly adheres to its own rules. The company prohibits using its models for autonomous weapon guidance, surveillance (including within the United States), and any actions that could cause harm.
In the summer of 2025, the U.S. Department of Defense entered into agreements with four leading AI companies—Anthropic, OpenAI, Google, and xAI. Each agreement could reach up to $200 million; their goal is the development of AI solutions for national security. The contract with Anthropic is for two years. However, due to the irreconcilable positions of the parties, the Pentagon is considering the possibility of terminating the contract. Replacing the Claude neural network will be difficult—it possesses technological advantages for solving specific government tasks.
The situation escalated after The Wall Street Journal published a report on the possible use of Claude in a U.S. forces special operation to capture Venezuelan President Nicolás Maduro. According to the publication, strikes were carried out on targets in Caracas during the operation. Anthropic declined to comment on its neural network's involvement in the operation but emphasized that any use of Claude must comply with the company's internal policy. At the same time, Anthropic's rules explicitly prohibit using AI for violence, weapons development, and organizing surveillance.
According to Axios, the Department of Defense insists that all four companies allow the use of their technologies "for all lawful purposes," including weapons development, intelligence gathering, and combat operations. The developers' reactions differ. One company has already agreed to the Pentagon's terms, two others show a willingness to compromise, while Anthropic maintains the toughest stance. OpenAI has already made concessions, granting access to its tools for 3 million defense department employees and lifting some restrictions on the use of ChatGPT.
This conflict exposes a fundamental contradiction of the modern era: on one hand, the military's drive to integrate cutting-edge AI technologies into combat operations; on the other, the ethical principles of developer companies, which fear the uncontrolled application of artificial intelligence.
The outcome of the negotiations remains uncertain. It will determine not only the fate of the $200 million contract but also how the relationship between the U.S. defense establishment and the AI industry will be structured moving forward.