Executive Order 14319, signed by President Donald Trump (47th) on July 23, 2025, establishes "Unbiased AI Principles" as a condition for federal procurement of large language models (LLMs). The order requires procured LLMs to be truth-seeking and ideologically neutral, and prohibits agencies from purchasing models that incorporate diversity, equity, and inclusion ("DEI") content as the order defines it. It was published on July 28, 2025 at 90 FR 35389.
Summary
The order conditions federal LLM procurement on compliance with "Unbiased AI Principles." It prohibits agencies from purchasing AI models that incorporate "DEI," which the order defines to include the suppression of factual information about race or sex, the manipulation of representation in outputs, and the incorporation of concepts such as critical race theory, transgenderism, or unconscious bias.
Unbiased AI Principles (Section 3)
Section 3 requires procured LLMs to comply with two principles:
- Truth-Seeking: LLMs shall be truthful, prioritizing historical accuracy, scientific inquiry, and objectivity. They shall acknowledge uncertainty where information is incomplete or contradictory.
- Ideological Neutrality: LLMs shall be neutral, nonpartisan tools that do not manipulate responses in favor of "ideological dogmas such as DEI." Developers shall not encode partisan or ideological judgments unless those judgments are "prompted by or otherwise readily accessible to the end user."
Implementation (Section 4)
Within 120 days, the Office of Management and Budget (OMB), in consultation with the Administrator for Federal Procurement Policy, the GSA Administrator, and the OSTP Director, must issue guidance to agencies. Per Section 4, that guidance must:
- Account for technical limitations
- Permit compliance via transparency disclosures (system prompt, specifications, evaluations) rather than requiring model weights
- Avoid over-prescription and allow different compliance approaches
- Specify factors for applying the Unbiased AI Principles to agency-developed models and non-LLM AI
- Make exceptions for national security systems
Agency heads must include Unbiased AI Principles compliance terms in new LLM contracts and, to the extent practicable, revise existing contracts.
Policy context
The order operationalizes the ideological framing of EO 14179 ("AI systems free from ideological bias or engineered social agendas") in the specific domain of federal procurement. It expresses the Trump administration's view that major AI labs have built ideological biases into their models — a premise partially supported by the Stanford GSB — Measuring Perceived Slant in LLMs (Westwood, Grimmer, Hall) and Manhattan Institute — Measuring Political Preferences in AI Systems (Rozado) studies, which independently found left-leaning biases in commercial LLMs.
The order does not regulate private-market AI; it applies only to federal procurement. Given the scale of federal AI contracts (GSA OneGov Program and USAi Platform (August 2025)), compliance with the Unbiased AI Principles could function as a de facto industry requirement for major labs.
Tension with safety frameworks
The order's "ideological neutrality" framing is in tension with AI safety frameworks that use harm-avoidance criteria. Refusing to generate detailed instructions for making weapons, for example, could be characterized as an "ideological" restriction. The order's exemptions for vendor transparency disclosures may provide a compliance pathway, but the line between safety refusals and ideological suppression is contestable.
Provenance
Signed July 23, 2025; published July 28, 2025 (90 FR 35389). Primary text held at Raw Sources/Executive Order 14319 - Preventing Woke AI in the Federal Government.md.
Relationships
- related: Executive Order 14179 — Removing Barriers to American Leadership in Artificial Intelligence — the first Trump AI EO; established the ideological framing this EO operationalizes
- related: Executive Order 14365 — the comprehensive positive AI policy framework
- related: AI Political Bias — the empirical studies that partially support this EO's premise
- related: GSA — General Services Administration (AI Deployer) — GSA is the primary federal AI procurement vehicle affected
- related: GSA OneGov Program and USAi Platform (August 2025) — GSA's OneGov program is the immediate procurement context