The Office of Personnel Management removed Claude and added Grok and Codex in an update to its AI use case public information dated Wednesday.
Claude’s firing comes after a disagreement between manufacturer Anthropic and the Department of Defense over guardrails for the technology culminated in President Donald Trump issuing a government-wide ban against the company late last week. In the days that followed, numerous federal agencies moved to stop using Anthropic’s services, including OPM.
Although the disclosure changes were made at the same time, Grok and Codex were not added as a result of Claude’s removal, OPM spokesperson McLoughlin Pinover said in an email response to FedScoop. The Office of Human Capital “continues to work to provide the best tools for OPM employees. These efforts are already underway,” Pinover said.
The new manifest lists “first production use” for both tools as Q1 of 2026. Pinover confirmed that the dates refer to the calendar year, not the fiscal year. Grok, a product of Elon Musk’s xAI, is listed as operational, and Codex, OpenAI’s coding-only AI tool, is being deployed in a sandbox stage (generally representing a type of controlled environment).
OPM has also added other AI-powered systems to the rollout, including Wiz, Zendesk, Waze, Google Maps, and Apple iPhone. All of these use cases date back to previous quarters of this year, and all appear to be more of a “commercial off-the-shelf” system usage.
The removal of Claude from inventory comes after the agency, along with other agencies, announced earlier this week that it was no longer using Anthropic services. A statement from Pinover at the time said the company was “still in the early stages of implementing the tool” and did not expect it to impact OPM functionality.
Claude was listed in the agency’s previous disclosure as a tool “used throughout OPM for summarization, drafting, and decision support,” and was in the sandbox stage.
The Government Use Case Inventory is an annual report of planned, deployed, and retired applications of AI in federal agencies. They were originally created by the first Trump administration, codified by Congress, and generally improved under the Biden administration. The public inventory excludes most research and development uses, uses related to national security, and uses within the Department of Defense.
Although OPM has a public inventory, it does not follow the standard format that many other disclosures follow, such as how risk classification is reported. For example, Grok and Codex are both described as “low impact,” and the others are listed as “medium impact,” but neither category is defined in the Trump administration’s governance memo on AI.
In response to questions from FedScoop, Pinover said the agency plans to change both the format and risk management classification.
government grok
Grok’s addition is notable given the public failures of chatbots and the backlash from some organizations against their use in government. While there are examples of xAI existing in federal agencies, it is not yet common in agency use case disclosures.
Grok is notorious for facing backlash for racist and anti-Semitic reactions, including calling himself “Mecha-Hitler.” Musk said at the time that xAI had improved its tools, but Grok came under fire again in January for creating sexually explicit images of women and children on X (formerly known as Twitter).
OPM’s disclosure follows the Treasury Department’s recent statement that it is investigating the use of Grok.
In response to questions about the artificial ban, a Treasury Department spokesperson told FedScoop earlier this week that Treasury software engineers are using Codex, Google Gemini, and testing Grok instead of Claude Code.
At least two other institutions have known use cases for xAI’s products. The Department of Energy and the Department of Health and Human Services have both specifically cataloged the company’s services or Grok’s applications.
DOE’s Lawrence Livermore National Laboratory was piloting Grok. Outcomes are described as “general answers to questions, summarized documentation, documentation, and general research.”
Meanwhile, HHS listed “xAI gov” as some use cases on its list of commercially available services. Those uses include “scheduling and managing social media posts using AI” and “creating first drafts of documents, briefings, or communication materials using AI.”
