The Office of Personnel Management (OPM) announced the removal of Claude from its public disclosure of artificial intelligence (AI) use cases, adding Grok and Codex in an update released Wednesday. This change follows a recent governmentwide ban on Claude’s developer, Anthropic, issued by President Donald Trump after a dispute over the technology’s safety protocols with the Department of Defense.
Numerous federal agencies, including OPM, have since halted the use of Anthropic’s services. OPM spokeswoman McLaurine Pinover clarified in an email response to FedScoop that the inclusion of Grok and Codex was not directly related to the removal of Claude, stating that the agency is “constantly working to provide the best tools to the OPM workforce” and that these initiatives were already in progress.
The updated inventory indicates that both Grok and Codex are expected to see their “first production use” in the first quarter of 2026, referring to the calendar year rather than the fiscal year. Grok, developed by Elon Musk’s xAI, is currently in production, while Codex, a coding-focused AI tool from OpenAI, is being deployed in a controlled sandbox phase.
In addition to these updates, OPM included several other AI systems in its disclosure, such as Wiz, Zendesk, Waze, Google Maps, and the Apple iPhone, backdating all these use cases to prior quarters. These applications are categorized as “commercial-off-the-shelf” products.
The removal of Claude comes after the agency reported earlier this week that it had ceased using any services from Anthropic, indicating that it was “still in the initial steps of implementing the tool” and did not anticipate any impact on OPM operations. Previously, Claude had been listed as a tool utilized across OPM for tasks like summarization, drafting, and decision support while still in a sandbox phase.
Federal agencies are required to maintain annual inventories of AI applications, detailing those that are planned, deployed, or retired. These disclosures were established during the Trump administration and later refined under President Biden. Notably, the public inventories generally exclude research-and-development uses, national security applications, and any related to the Department of Defense.
Although OPM maintains a public inventory, it has drawn criticism for not adhering to standard reporting formats used by other agencies, particularly regarding risk classifications. For instance, both Grok and Codex are categorized as “Low-impact,” while other tools are labeled as “Medium-impact,” yet there is no definition for these categories in the Trump administration’s AI governance memo. In response to inquiries, Pinover stated that OPM plans to revise both the format and the classifications for risk management.
Grok’s Controversial History
The addition of Grok is particularly noteworthy given its history of public missteps and criticism from various organizations regarding its potential use in government. While xAI’s presence in federal agency reports is limited, Grok has notably faced backlash for producing racist and antisemitic responses, including the controversial self-reference “MechaHitler.” Despite Musk’s assertions that improvements were made to the tool, Grok encountered further criticism earlier this year for generating inappropriate images on X, formerly known as Twitter.
OPM’s recent disclosure aligns with a statement from the Department of the Treasury, which is also exploring the use of Grok. A spokesperson for the Treasury noted that software engineers are currently using Codex, Google Gemini, and testing Grok as alternatives to Claude Code.
At least two additional federal agencies have also reported applications for xAI’s products. The Department of Energy (DOE) has been piloting Grok through its Lawrence Livermore National Laboratory, which described its outputs as providing general answers to questions, summarizing documents, and supporting general research. Meanwhile, the Department of Health and Human Services (HHS) has included “xAI gov” in its inventory for tasks such as scheduling and managing social media posts, in addition to generating first drafts of documents and communications materials using AI.
See also
Germany”s National Team Prepares for World Cup Qualifiers with Disco Atmosphere
95% of AI Projects Fail in Companies According to MIT
AI in Food & Beverages Market to Surge from $11.08B to $263.80B by 2032
Satya Nadella Supports OpenAI’s $100B Revenue Goal, Highlights AI Funding Needs
Wall Street Recovers from Early Loss as Nvidia Surges 1.8% Amid Market Volatility
