By G. Murphy Donovan
“Before we work on artificial intelligence why don’t we do something about natural stupidity?” — Steve Polyak
The Pentagon vs. Anthropic dispute (escalating rapidly in late February 2026 under the Trump administration) centers on a contract for Anthropic’s Claude AI model, deployed on classified U.S. military networks (part of up to $200M deals awarded in 2025 to several AI firms).
Key issues:
- The Pentagon (led by Defense Secretary Pete Hegseth) demanded Anthropic remove all safeguards/restrictions on Claude’s use, insisting on access for any lawful military purpose without company-imposed limits.
- Anthropic (CEO Dario Amodei) refused, citing red lines against:
-
- Fully autonomous weapons (lethal systems selecting/engaging targets without human oversight, due to current reliability/safety risks).
- Mass surveillance of U.S. citizens (domestic, without proper oversight).
- After a Friday (Feb 27, 2026) deadline passed without compliance, the administration escalated:
-
- President Trump ordered all federal agencies to stop using Anthropic tech (with a 6-month DoD phaseout).
- The Pentagon designated Anthropic a “supply chain risk to national security” — an unprecedented label for a U.S. company (typically for foreign adversaries), barring military contractors from doing business with it.
- Anthropic called the moves legally unsound, vowed to challenge the designation in court, and expressed being “deeply saddened” by the rift.
- Meanwhile, rival OpenAI quickly announced a deal with the DoD for its tech on classified systems.
In short, it’s a high-stakes clash over who controls AI guardrails and national security: company ethics/safety policies vs. government demands for unrestricted “lawful” use — with major implications for AI-military integration, precedents for U.S. firms, and the balance of power between Big Tech and the DOD establishment.
Open AI and Elon Musk’s Xai appear ready and willing to fill any vacuum created by the departure of Anthropic at DOD. Ironically, this clash of titians in DC is just an echo of the earlier clash between Elon Musk and the original founders of Open AI.
In that battle, “open and safe” lost to closed and profitable – and Elon Musk’s subsequent decision to create Xai and Grok.
Who controls Artificial Intelligence engines will have a lot to say about who controls many things including potent weapons and policy narratives. To be continued….
Hat tip to Grok.
- Murphy Donovan writes about the politics of Intelligence and national security.
Illustration BigC studio
2 Responses
My observation would be that control of the the AI guardrails seems to be a repeat of the disagreement currently between the Deep State and the Trump Admnistration. Would the ‘expert’ creators of the program [regulation] control it, or would our military [Trump]?
Yeah, Carl. At the moment it’s a toss up. If the original open AI fracas is precedent, I’d say pay walls and secrecy prevail. What can be monetized, will be monetized, all thru a glass darkly.