The Department of Defense is actively evaluating AI models from OpenAI, Google, and xAI’s Grok as replacements for Anthropic’s Claude, which was effectively blacklisted from military use in late February. The competitive testing kicked off on March 1, 2026, just days after Defense Secretary Pete Hegseth designated Anthropic’s products as a supply-chain risk.

The core issue: Anthropic refused to remove safety restrictions that prevent its AI from being used for mass surveillance or lethal autonomous weaponry.

A $200 million contract unravels

Anthropic secured a contract worth up to $200 million in July 2025 to integrate Claude into classified military networks.

Things started falling apart when Pentagon officials pushed for unrestricted use of the technology, particularly in areas related to national security and domestic surveillance. Anthropic’s leadership drew a line. The Pentagon drew a different one.

The standoff escalated publicly in January 2026 when reports emerged that Claude had been used for intelligence analysis and operational planning related to Iran.

By late February, Hegseth had seen enough. His February 27 designation of Anthropic as a supply-chain risk set the wheels in motion for a full replacement.

25 military testers, one new platform

The evaluation is running through GenAI.mil, a platform that operates independently of the Pentagon’s existing Maven Smart System. Twenty-five designated military personnel are conducting the tests, putting competing models through their paces in workflows that previously relied on Claude.

The models under scrutiny include offerings from OpenAI, Google, and Grok from Elon Musk’s xAI.

Contractors who had integrated Anthropic’s technology into their systems were given a six-month window to find alternatives.

Anthropic, for its part, has filed lawsuits over the situation.

The bigger picture for defense AI

OpenAI and other tech entities are already moving to secure new classified contracts with the Department of Defense.

Anthropic was founded specifically as a safety-focused alternative to OpenAI. Its entire corporate identity rests on the idea that AI development should come with meaningful restrictions. Being told those restrictions are a dealbreaker for a $200 million military contract is essentially being told that the company’s core value proposition is incompatible with its most lucrative customer.