Descriptions:
Bloomberg Technology reports on a consequential policy standoff between Anthropic and the US Department of Defense: Anthropic has loosened its central safety policy amid competitive pressure and a Pentagon ultimatum with a hard Friday deadline. Bloomberg senior tech editor Mike Shepherd breaks down the specific legal threats and conditions involved.
The dispute centers on two conditions Anthropic had attached to any government use of its technology — prohibiting mass surveillance of Americans and preventing fully autonomous weapons targeting. The Pentagon’s new AI acceleration strategy explicitly rejects any vendor conditions on government use of technology. Faced with non-compliance, the Defense Department threatened to invoke the Defense Production Act to compel access to Anthropic’s models, or alternatively to declare Anthropic a formal supply chain risk — a designation that would require all Pentagon vendors to certify they are not using Anthropic’s products, posing a severe commercial threat to the company’s enterprise business.
CEO Dario Amodei met with Defense Secretary Pete Hegseth ahead of the deadline. Bloomberg reports that while Anthropic has adjusted its broader safety policy in response to competitive dynamics — acknowledging it faces stiff pressure from OpenAI, xAI, and Google — it is holding firm on the two core conditions around surveillance and autonomous weapons. The episode highlights the deepening tension between AI labs’ publicly stated safety commitments and the commercial realities of operating in a market where the US government is both a major potential customer and a regulatory actor with significant coercive power.
📺 Source: Bloomberg Technology · Published February 25, 2026
🏷️ Format: News Analysis
