The Pentagon has severed ties with AI developer Anthropic after a failed $200 million deal, abruptly ending a major government-AI partnership, forcing the Department of War into a rapid rethink.

The dispute centered on limits on surveillance and the use of autonomous weapons, with the core issue being how far the military should be allowed to use AI in sensitive operations.

Anthropic pushed strict safeguards prohibiting its AI from being used for systemic public monitoring or for automating weapons, but the Pentagon rejected those key conditions, calling them too restrictive.

After negotiations broke down quickly, defense officials pivoted rapidly to alternative AI providers such as OpenAI and Grok.

Anthropic is now preparing a legal challenge against the “supply-chain risk” label given by War Secretary Pete Hegseth, in effect fighting a label that could damage its future government business.

By-products of the environment

The AI pack has been shuffled somewhat. Amid the chaos, Sam Altman's OpenAI has secured a new Pentagon contract, and Altman recently admitted that the firm cannot control the military's AI use. OpenAI’s chief of robotics subsequently quit her position in a decision of conscience following the deal.

Elon Musk’s xAI has also offered the Grok model for classified use, meaning the AI will operate behind Pentagon firewalls to support tasks such as research, analysis, and decision-making.

This fragmentation may cause confusion and disarray within government agencies. Certain branches, such as the General Services Administration and the Department of Health and Human Services, had to abruptly pull the plug on Anthropic's Claude after Donald Trump issued an order to ban the AI.

Thousands of users lost access to chats, code, and workflows instantly, with ongoing work and saved data disappearing overnight.

Some agencies are still running the old Anthropic AI systems because they haven’t been told how to switch safely.

The Cybernews community is talking about this. Be a part of the conversation.

This is all despite reports that the US army used Claude anyway during attacks on Iran, soon after the ban.

Anthropic and The Pentagon have been involved in a running stand-off of late, with speculation rife from both sides, as well as with leading researchers regarding autonomous weapons and homeland surveillance capabilities.