Pentagon Pulls the Plug on Anthropic AI After Earlier Blacklisting

The Pentagon’s AI policy has taken a curious turn. Earlier this year, the Department of Defense officially labeled Anthropic a "supply‑chain risk," effectively blacklisting the company’s products.

Months after that designation, defense officials announced that the Pentagon had stopped using Anthropic’s Claude model. The decision came even though multiple sources said Claude had been widely used in recent briefings and simulations.

The move highlights a tension within the department: a tool deemed risky enough to be blacklisted was still being relied upon for its capabilities, only to be pulled back once the policy caught up with practice.

Analysts note the abrupt shift underscores how quickly the DoD’s AI guidelines can evolve, and how the line between "useful" and "risky" can blur in a fast‑moving technology landscape.

The episode serves as a reminder that even in the world of defense tech, policy changes can be as swift and surprising as the technologies they aim to govern.