Pentagon Weighs Terminating Anthropic Claude Contract Amid Ethical Dispute

by Park SeungUk

Published 15 Feb.2026 16:45(KST)

In talks with OpenAI, Google, xAI, and others

The political news outlet Axios reported on the 14th (local time) that the U.S. Department of Defense is considering terminating its contract with Anthropic, the developer of the artificial intelligence (AI) system Claude that was used in the operation to arrest and extradite Venezuelan President Nicolas Maduro.

Anthropic logo. Reuters Yonhap News

The U.S. Department of Defense and Anthropic are at odds over ethical issues surrounding the use of AI on the battlefield. According to the report, the Pentagon has been negotiating with Anthropic for several months to use AI legally in sensitive areas such as weapons development, intelligence gathering, and battlefield operations, but progress has been slow.

Anthropic maintains that Claude must not be used for all-encompassing surveillance of U.S. citizens or for fully autonomous weapons systems.

Anthropic, founded by former OpenAI employees who were critical of commercialization, has championed "safe and ethical AI" and has maintained guidelines that restrict its models from being used directly in the development of lethal weapons or in military operations. However, tensions between the Pentagon and Anthropic are reported to have escalated after the U.S. military used Claude last month in Operation Absolute Resolve, the mission to arrest and extradite Maduro.

In this regard, the U.S. Department of Defense takes the position that there is a significant gray area within the prohibited domains asserted by Anthropic. It argues that it is not feasible to negotiate with Anthropic on a case-by-case basis for every individual instance, nor is it realistic to allow Claude to unexpectedly block specific applications (apps).

From the Pentagon’s perspective, it is difficult to replace Anthropic’s Claude, which is currently being used on classified networks, but it is also actively negotiating with other AI companies. OpenAI, Google, and xAI are reportedly under consideration as negotiation counterparts; one of them has accepted the information request, and the other two are said to be taking a more flexible stance than Anthropic.

Claude by Anthropic is the only AI currently used on the U.S. Department of Defense’s classified networks. The Pentagon signed a software licensing agreement worth 200 million dollars (about 290 billion won) with Anthropic last summer.