Executive Order 14319, "Preventing Woke AI in the Federal Government," was signed by President Donald Trump on July 23, 2025, and published July 28, 2025 (90 FR 35389). It conditions federal procurement of large language models (LLMs) on two "Unbiased AI Principles" — truth-seeking and ideological neutrality — and directs the Office of Management and Budget (OMB) to issue implementing guidance to agencies within 120 days.
| Signed | July 23, 2025 |
| Published | July 28, 2025 (90 FR 35389) |
| President | Donald Trump (47th) |
| Scope | Federal LLM procurement; all executive agencies |
Status and timeline
The order was signed on July 23, 2025 and published in the Federal Register on July 28, 2025. It directed OMB to issue agency guidance within 120 days of signing, placing the deadline at approximately November 20, 2025. Going forward, new federal LLM contracts must include compliance terms tied to the Unbiased AI Principles, and existing contracts are to be revised where practicable. OMB issued an implementation memo in December 2025 (see Compliance pathway).
Unbiased AI Principles
The order permits federal agencies to procure only LLMs developed in accordance with two principles:
- Truth-Seeking: historical accuracy, scientific inquiry, and objectivity, including acknowledgment of uncertainty where information is incomplete or contradictory.
- Ideological Neutrality: no manipulation in favor of "ideological dogmas such as DEI," and no encoding of partisan or ideological judgments unless user-prompted.
Definition of "DEI"
The order defines DEI in the AI context as: suppression or distortion of factual information about race or sex; manipulation of racial or sexual representation in model outputs; incorporation of concepts such as critical race theory, transgenderism, unconscious bias, intersectionality, or systemic racism; and discrimination on the basis of race or sex. Under this definition, content restrictions, refusals, or representation choices that fall under these categories would make an LLM non-compliant with the Unbiased AI Principles for federal procurement.
Compliance pathway
The order permits compliance through transparency disclosures — system prompts, specifications, and evaluations — rather than requiring disclosure of model weights. This provides a route for major labs to document and disclose their design choices rather than fundamentally redesigning models. The order states that disclosure can serve as a compliance route but does not mandate it.
Neumann, Sargeant & Singh (FAccT'26) use EO 14319 as a case study of what they term a "compliance illusion." They note that the December 2025 OMB implementation memo lets vendors evidence "ideological neutrality" via system-prompt disclosure but excludes system prompts from the memo's "model evaluations" section, embedding the assumption that inspecting prompt language is sufficient to demonstrate neutral behavior. On their account, a carefully drafted system prompt may satisfy the disclosure requirement while the model's actual outputs remain unverified (see Prompt Governance).
Interaction with federal AI procurement
GSA OneGov Program and USAi Platform (August 2025) documents the GSA OneGov/USAi platform, which negotiates AI contracts on behalf of all federal agencies at $1/agency pricing. EO 14319's requirements would apply to all models deployed through this platform, making the Unbiased AI Principles a procurement standard across the federal AI stack.
Areas of tension
Commentators and the order's own text identify several points of friction in applying the Unbiased AI Principles:
- Safety refusals and ideological neutrality. Models that refuse harmful content requests could be characterized as ideologically suppressing information. The technical-limitations exemption in Sec. 4(a)(i) provides some buffer.
- Vendor transparency burden. Documenting system prompts and specifications may expose competitive intellectual property. The order says disclosure can be a compliance route but does not mandate it.
- State and federal divergence. California SB 53 and the New York RAISE Act require safety disclosures that implicitly assume some content restrictions are appropriate, whereas EO 14319 pushes federal procurement in the opposite direction.
- Verification of prompt-based neutrality. Per Neumann, Sargeant & Singh, the system-prompt-disclosure pathway may certify the language of a prompt without verifying the model's behavior (see Compliance pathway and Prompt Governance).
Relationships
- related: Executive Order 14179 — Removing Barriers to American Leadership in Artificial Intelligence — the January 2025 EO that established the ideological framing
- related: Executive Order 14365 — Ensuring a National Policy Framework for AI — comprehensive positive AI policy framework
- related: AI Political Bias — empirical studies on LLM bias that provide evidentiary basis
- related: GSA — General Services Administration (AI Deployer) — primary federal procurement agency affected
- related: California SB 53 — state-level transparency requirement; points in different direction
- related: Prompt Governance, Prompt Governance? On Governing Technologies Governed by Natural Language — critique of the system-prompt-disclosure compliance pathway