StateChat users move to older AI model as State dumps Anthropic

Report Content

The State Department has shifted the model underpinning its internal chatbot, StateChat, from Anthropic’s Claude Sonnet 4.5 to OpenAI’s GPT-4.1, according to an internal document obtained by Nextgov/FCW.

The document also indicates that less data is available to StateChat users, at least temporarily. The new model has only been trained on data through May 2024. The Claude-powered version of StateChat was trained on data as recent as June 2025, a source familiar with the situation told Nextgov/FCW. The person spoke on the condition of anonymity to provide details about the model’s training data.

The document also shows that State employees using a customGPT setup running on Claude were asked to move to a non-Anthropic tool by March 6.

The switch comes after President Trump's Feb. 27 directive ordering government agencies to evict Anthropic tools from their systems. 

“In line with the president’s direction to cancel Anthropic contracts, Anthropic’s Claude models are no longer available on the Department’s enterprise generative AI platform,” a State Department spokesperson told Nextgov/FCW. “The department is taking all necessary steps to implement the directive and bring our programs into full compliance.”

Reuters previously reported directives in multiple agencies, including State, requiring switches from Claude to ChatGPT.

Claude was initially made available for federal agency operations as part of the General Services Administration’s OneGov deal that brokered favorable software rates for the government, many for a temporary period of time.  

Following the governmentwide ban of Anthropic technology, the company filed two lawsuits Monday. One, filed in the D.C. circuit court, invokes provisions in the Federal Acquisition Supply Chain Security Act of 2018 as rationale against the government’s designation of Anthropic as a supply chain risk.

The second, filed in the U.S. District Court in the Northern District of California, names dozens of federal agencies and officials as defendants in allegations of inappropriate retaliation against Anthropic and calls for an injunction. 

Classification
Region
North America
Analytical Domain
Strategic
Primary Category / Secondary Categories
Political-Military / Information Operations
SALUTE Report
Activity
The State Department has shifted its internal chatbot model from Anthropic’s Claude Sonnet 4.5 to OpenAI’s GPT-4.1, following a directive from President Trump to evict Anthropic tools from government systems.
Location
Washington, D.C.
Unit
U.S. State Department
Time
March 6, 2024
Equipment
OpenAI's GPT-4.1Anthropic's Claude Sonnet 4.5
Summary

The U.S. State Department shifted its internal chatbot model from Anthropic’s Claude Sonnet 4.5 to OpenAI’s GPT-4.1 by March 6, 2024, following President Trump's directive to remove Anthropic tools from government systems. This transition is part of compliance efforts to align with the directive, which has led to Anthropic filing lawsuits against the government for alleged retaliation. The new model has limited training data compared to its predecessor.

Key Facts
  • The State Department has switched its chatbot model from Anthropic’s Claude Sonnet 4.5 to OpenAI’s GPT-4.1.
  • This change follows a directive from President Trump to remove Anthropic tools from government systems.
  • State employees were instructed to transition to non-Anthropic tools by March 6, 2024.
  • The Claude-powered version of StateChat was trained on data as recent as June 2025, while the new model is trained only through May 2024.
  • Anthropic has filed lawsuits against the government following the ban of its technology.