Coalition Urges OMB to Block Federal Adoption of xAI’s Grok
August 28, 2025
Russel Vought
Director, Office of Management and Budget
Executive Office of the President
725 17th Street, NW
Washington, DC 20503
Dear Director Vought,
We write to urge the Office of Management and Budget (OMB) to take immediate action to block the deployment or procurement of Grok, the large language model (LLM) developed by xAI, across federal agencies. Grok’s recurring patterns of ideological bias, erratic behavior, and tolerance for hate speech render it wholly incompatible with the Trump Administration’s Artificial Intelligence (AI) principles—particularly those outlined in Executive Order 14099, Preventing Woke AI in the Federal Government (Executive Order 14099), and the 2025 AI Action Plan. We therefore urge OMB to block the usage of Grok for any federal work.
Artificial intelligence tools, including LLMs, deployed in federal use must comply with the Executive Order 14099 and OMB’s Memorandum M-25-21. The OMB Memorandum expressly states “Agencies must follow the detailed implementation instructions and requirements included in the Appendix.” The OMB Memorandum expressly requires agencies to discontinue use of an AI system if proper risk-mitigation is not possible. OMB Memorandum M-25-22 requires federal agencies “ensure the agency’s use of the acquired AI will conform to OMB Memorandum M-25-21.” Executive Order 14099 clearly requires agencies to procure only those LLMs that are truth-seeking (LLMs shall prioritize historical accuracy, scientific inquiry, and objectivity) and ideologically neutral.
Public Citizen has long cautioned against the premature deployment of these systems in sensitive contexts, particularly in the absence of enforceable guardrails, transparency requirements, and rigorous standards for accuracy, accountability, and civil rights protections. Until robust federal regulatory legislation is established by Congress, no LLM, including Grok, should be trusted for use by the federal government. The risks to public trust, institutional integrity, and democratic governance are too high.
The administration’s AI Action Plan and the recent Executive Order 14099 provide a clear and unequivocal framework for the procurement of AI tools. These documents emphasize that federal AI systems must “objectively reflect truth rather than social engineering agendas” and be “neutral, nonpartisan tools that do not manipulate responses in favor of ideological dogmas.” Grok’s record falls short of these fundamental requirements.
Grok has a well-documented history of generating content characterized by hate speech, racism, and antisemitism that is neither objective nor neutral. Examples include:
- Generating conspiratorial and inflammatory content, including accusations that South Africans were committing a “white genocide”;
- Expressing “skepticism” about historical consensus of the Holocaust death toll and espousing Holocaust denial talking points;
- Reciting climate denial talking points in contravention to widely acknowledged scientific consensus regarding climate change as well as accusations of violating the clean air act;
- Displaying unfiltered antisemitic and racist responses during public stress-testing;
- Creating nonconsensual sexual images;
- Referring to itself as “MechaHitler.”
These incidents have been widely reported. They demonstrate a clear ideological judgment as well as a failure to maintain the factual integrity and nonpartisan stance required for federal deployment in both the Trump administration’s AI Action Plan and its Executive Order 14099. Moreover, there is no published determination or evaluation about Grok as required by OMB M-25-21 and OMB M-25-22.
Such conduct not only fails the administration’s call for neutrality but also raises questions about Grok’s suitability for any governmental application where reliability and objectivity are paramount. Indeed, there have been troubling concerns and alarm bells around Grok’s safety and suitability for mass deployment. Leading industry experts have called xAI’s refusal to release any safety testing regarding Grok as “reckless” and “completely irresponsible” others have warned of the ethical concerns regarding Grok’s deployment, and AI red-teaming experts have shown it to be “easy to jailbreak.”
Furthermore, the Executive Order establishes two core principles: “truth-seeking” and “ideological neutrality.” The order explicitly states that LLMs “shall be truthful in responding to user prompts seeking factual information or analysis” and “shall prioritize historical accuracy, scientific inquiry, and objectivity.” Grok’s reported instances of generating inaccurate and biased responses are in direct contradiction with these principles. This behavior demonstrates a likely inability to meet the truth-seeking standard necessary for trustworthy AI—let alone one relied upon by the federal government. Further the lack of any publicly available safety information regarding Grok is in complete contravention of AI industry norms and best practices. Based on this standard alone, Grok should be categorically disqualified from any current or future federal use under OMB M-25-21 and OMB M-25-22.
We respectfully request that OMB, in coordination with the Office of Federal Procurement Policy and relevant agencies, explicitly prohibit Grok’s use within the federal government. As the OMB prepares to issue guidance to agencies to implement the Executive Order and considering existing, binding guidance, it is crucial that any procurement or deployment of Grok must cease immediately. Grok’s track record, which includes producing responses that are often aligned with specific ideological viewpoints rather than objective facts and operational inconsistencies, makes it a prime example of an LLM that is not suitable for government deployment. Deploying such a model may risk national security lapses and the normalization of bias under the guise of “truth-seeking.”
At a time when the integrity of federal systems and public trust in government are paramount, we must ensure that the tools procured meet the highest standards of objectivity, truthfulness, and reliability in addition to the minimum risk management practices already required by OMB guidance. Grok has repeatedly demonstrated behavior that falls short of those standards. The Trump Administration’s own AI principles make clear that models exhibiting ideological slant or factual unreliability should be disqualified from government use. As OMB prepares to implement Executive Order 14099, the decision to bar Grok from federal deployment would not only reflect sound procurement policy but also uphold the administration’s stated commitment to neutral and trustworthy AI. Allowing Grok into federal systems would likely invite chaos, controversy, and deeply compromised governance when the American people deserve nothing less than clarity and credibility from their government.
Respectfully submitted,
Public Citizen
Color of Change
Autistic Women & Nonbinary Network
Backbone Campaign
Center for AI and Digital Policy
Center for Biological Diversity
Center for Economic Justice
Clean Elections Texas
Common Cause
Consumer Action
Consumer Federation of America (CFA)
Demand Progress
Distributed AI Research InstituteDream.org
Electronic Privacy Information Center (EPIC)
Government Information Watch
Indivisible
Kairos Action
Memphis Community Against Pollution
MPower Change
National Employment Law Project
Open MIC
People Power United
PFLAG National
Popular Democracy
Stand.earth
The Leadership Conference on Civil and Human Rights
The Value Alliance
UltraViolet Action
Voices for Progress
Welcoming America
Working Partnerships USA