The Trump administration is closing the loop on a classified contract that will let US intelligence agencies, the NSA chief among them, tap into Anthropic’s AI models. The deal includes access to Anthropic’s newly developed Mythos model, with one critical guardrail: the technology cannot be used on data pertaining to US citizens.

From blacklisted to back in business

In February 2026, President Trump issued an order for an immediate cessation of all federal use of Anthropic’s technology. The reason was straightforward: Anthropic refused to give the Pentagon unrestricted access to its AI systems. The Pentagon designated Anthropic as a supply chain risk.

Anthropic’s Claude AI models had already been deployed across US government networks at that point. The company had received over $8 billion in funding from Amazon. Cutting it off created real operational gaps.

Those gaps got worse thanks to ongoing chip shortages that have constrained spy agencies’ ability to effectively run advanced AI models. White House Chief of Staff Susie Wiles authorized the NSA to continue using Anthropic’s models on an interim basis while negotiations over a new contract played out.

The deal takes shape

By April 2026, Trump described a potential deal with Anthropic for Department of Defense use, claiming the company was “shaping up” to meet government expectations.

The White House has committed $9 billion to enhance AI infrastructure specifically for defense agencies. The carve-out prohibiting use of Anthropic’s technology on US citizen data appears to be the compromise that got both sides to the table.