U.S. government officially began a sweeping transition away from Anthropic’s Claude, switching several key federal agencies to OpenAI’s GPT-4.1.
This move follows a dramatic executive order from President Trump and a directive from Defense Secretary Pete Hegseth, who labeled Anthropic a “supply chain risk” after the company refused to remove safety guardrails for military use.
The “GPT-4.1” Transition
The State Department has already moved its in-house chatbot, StateChat, from Claude to GPT-4.1. Other agencies currently phasing out Anthropic products in favor of OpenAI and Google Gemini include:
- Department of the Treasury
- Department of Health and Human Services (HHS)
- Federal Housing Finance Agency (FHFA)
- Fannie Mae & Freddie Mac
Why the Government Dumped Anthropic
The fallout stems from a high-stakes standoff between Anthropic CEO Dario Amodei and the Department of War (formerly the DoD). Anthropic insisted on two “red lines” in its government contracts:
- No Mass Domestic Surveillance: Prohibiting the use of AI to assemble fragmented data into comprehensive profiles of American citizens.
- No Fully Autonomous Weapons: Refusal to allow AI to make final lethal targeting decisions without a human in the loop.
Secretary Hegseth described Anthropic’s refusal to allow “unrestricted access for all lawful purposes” as an act of “corporate virtue-signaling” and “arrogance,” leading to the unprecedented supply-chain risk designation typically reserved for foreign adversaries like Huawei.
OpenAI’s Role and “GPT-4.1” Clarification
As Anthropic was blacklisted, OpenAI quickly moved to fill the vacuum. On February 27, OpenAI announced a new deal to deploy its models on the Department of War’s classified networks.
- What is GPT-4.1? While newer models like GPT-5.2 are now standard for consumers, GPT-4.1 is being utilized for specific government applications like StateChat due to its stability and specialized fine-tuning for administrative and intelligence tasks.
- The “Loophole” Controversy: Unlike Anthropic, OpenAI accepted the government’s “all lawful purposes” framework. While OpenAI claims it still has internal red lines against autonomous weapons, critics argue their contract lacks the explicit legal prohibitions that Anthropic demanded, particularly regarding the use of “publicly available information” for surveillance.
The Six-Month Phase-Out
The government has not disconnected Anthropic immediately. To avoid a “dark period” in military operations, a six-month transition window has been granted for the Department of War to migrate its existing embedded systems to OpenAI or other “patriotic” providers.
