Pentagon tests AI models to find alternatives to Anthropic’s Claude

The Pentagon is actively evaluating AI models from multiple providers as it moves to replace Anthropic’s Claude across its operations.

From partner to supply chain risk

In July 2025, Anthropic secured a two-year, $200 million contract to integrate Claude into classified defense networks.

Things started to unravel in January 2026. Claude was reportedly deployed during a military operation connected to the capture of Venezuelan President Nicolás Maduro. Anthropic’s contractual stipulations around “lawful use” apparently clashed with how the technology was being applied in that context, creating what sources describe as significant friction with the Department of Defense.

Anthropic had drawn clear lines about what it would and wouldn’t support. Fully autonomous weapons systems and extensive surveillance measures fell on the “wouldn’t” side.

By late February 2026, Defense Secretary Pete Hegseth issued an ultimatum. When Anthropic held its ground, Hegseth designated the company a “supply chain risk.” The designation triggered a phased transition away from Claude, with an expected timeline of three to twelve months for the full changeover.

OpenAI and xAI step into the gap

The Pentagon has fast-tracked alternative AI solutions from OpenAI and Elon Musk’s xAI, maker of Grok, on its GenAI.mil platform for military use.

OpenAI renegotiated its agreement with the DoD to ensure unrestricted lawful use of its AI systems, a direct contrast with the terms Anthropic refused to accept.

In a twist that captures the messiness of the whole situation, the NSA was internally testing Anthropic’s Claude Mythos preview on classified networks as recently as April 2026, even as the broader Pentagon was actively trying to phase Claude out.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.