Pentagon Moves to Build Its Own AI Alternatives After Deal With Anthropic Collapses

The Defense Department has begun work to replace Anthropic’s artificial-intelligence tools with systems hosted inside government networks, a senior Pentagon official said, signaling a clear break in the short-lived procurement relationship. Engineering teams have started adapting multiple advanced conversational models for use on classified and other controlled platforms, and the department expects at least some of those systems to be ready for operational testing in the near term, the official added.

The split follows the collapse of a roughly $200 million agreement after negotiations faltered over how the military could use the company’s software. Anthropic pushed to prohibit the department from using its technology for broad surveillance of citizens or for weapons that operate without human oversight. Defense leaders balked at contractual limits on operational access, and within weeks the Pentagon reached separate arrangements with other vendors, including one with OpenAI and another to integrate Grok from xAI into certain secure systems.

Officials say the push to stand up homegrown alternatives is intended to reduce dependence on any single supplier and to ensure sensitive workloads run inside government-controlled environments. “This is about keeping mission-critical capabilities where the department can maintain direct custody and control,” said a defense technology analyst who follows procurement trends. The analyst noted, however, that building and accrediting sophisticated language systems inside the government will be costly and time-consuming and may strain already tight acquisition schedules.

The department has also taken an aggressive regulatory step: Defense Secretary Pete Hegseth labeled Anthropic a supply-chain risk, a designation that can block companies that contract with the Pentagon from doing business with the AI firm. Anthropic has filed a legal challenge to that designation and vowed to contest the ruling in court, adding a legal layer to what had been a largely commercial dispute.

For military users and private contractors, the shift could mean retooling existing workflows and rewriting integration plans. Some industry observers see the move as likely to fragment the market for defense AI, while others argue it could spur investment in government-operated capabilities. Either way, officials appear to be preparing to move forward without Anthropic as they adapt their tools for sensitive operations.