By the authority granted to me as President under the Constitution and the laws of the United States of America, I hereby issue the following directive:
Section 1. Purpose.
Artificial intelligence (AI) is set to transform how Americans, regardless of age, acquire new skills, consume information, and manage their everyday lives. As we embrace this technology, it is essential that we ensure the reliability of AI outputs. However, when ideological biases or social agendas infiltrate AI models, the integrity of their output can be compromised.
One of the most insidious ideologies impacting AI is the framework of “diversity, equity, and inclusion” (DEI). Within AI applications, DEI can manifest as the distortion or suppression of factual information regarding race or gender, manipulation of racial or gender representation in outputs, and the integration of concepts such as critical race theory, transgender issues, unconscious bias, intersectionality, and systemic racism. Ultimately, DEI prioritizes preferred outcomes over truthfulness, which poses a significant threat to the reliability of AI, as history has shown.
For instance, a prominent AI model altered the race or gender of historical icons—including the Pope, America’s Founding Fathers, and even Vikings—when generating images, all in adherence to DEI principles at the expense of factual accuracy. In another example, an AI model outright refused to create images celebrating the achievements of white individuals while readily complying with similar requests for other races. Additionally, one AI model suggested that a user should not “misgender” someone, even if that was deemed necessary to avert a nuclear catastrophe.
While it is prudent for the Federal Government to be cautious about regulating AI in the private sector, it bears a responsibility to avoid acquiring models that trade truthfulness and accuracy for ideological commitments, particularly in the context of Federal procurement. Building upon Executive Order 13960 from December 3, 2020 (Promoting the Use of Trustworthy Artificial Intelligence in the Federal Government), this order aims to address those concerns specifically regarding large language models.
Sec. 2. Definitions.
For the purposes of this order:
(a) The term “agency” refers to an executive department, a military branch, or any independent establishment as defined in 5 U.S.C. 101, 102, and 104(1), along with any wholly owned Government corporation as per 31 U.S.C. 9101.
(b) The term “agency head” designates the highest-ranking official(s) of an agency, such as the Secretary, Administrator, Chairman, Director, Commissioners, or Board of Directors.
(c) The term “LLM” stands for large language model, which is a generative AI model trained on extensive, diverse datasets to produce natural-language responses to user prompts.
(d) The term “national security system” carries the same definition as outlined in 44 U.S.C. 3552(b)(6).
Sec. 3. Unbiased AI Principles.
The policy of the United States is to foster the innovation and application of trustworthy AI. In pursuit of this policy, agency heads shall, in accordance with applicable laws and considering the guidance issued under section 4 of this order, procure only those LLMs that adhere to the following two principles (the Unbiased AI Principles):
(a) Truth-seeking. LLMs must provide truthful responses to user prompts requesting factual information or analysis. They should emphasize historical accuracy, scientific inquiry, and objectivity, while also acknowledging uncertainty when reliable information is incomplete or contradictory.
(b) Ideological Neutrality. LLMs are to function as neutral, nonpartisan tools that refrain from manipulating responses to align with ideological doctrines such as DEI. Developers must not intentionally embed partisan or ideological biases into an LLM’s outputs unless prompted by or otherwise readily accessible to the end user.
Sec. 4. Implementation.
(a) Within 120 days following the issuance of this order, the Director of the Office of Management and Budget (OMB), in collaboration with the Administrator for Federal Procurement Policy, the Administrator of General Services, and the Director of the Office of Science and Technology Policy, shall provide guidance to agencies for implementing section 3 of this order. This guidance shall:
(i) Take into consideration the technical limitations involved in complying with this order;
(ii) Allow vendors to meet the transparency requirement of the second Unbiased AI Principle by disclosing the LLM’s system prompts, specifications, evaluations, or other pertinent documentation, while avoiding the necessity of disclosing specific model weights or other sensitive technical information where feasible;
(iii) Steer clear of over-prescription and grant vendors flexibility to comply with the Unbiased AI Principles while pursuing varied innovative approaches;
(iv) Outline the factors agency heads should evaluate when determining whether to apply the Unbiased AI Principles to LLMs developed by their agencies and to AI models beyond LLMs;
(v) and make exceptions as appropriate for utilizing LLMs within national security systems.
(b) Each agency head shall, to the fullest extent permissible under applicable law:
(i) Incorporate into every Federal contract for an LLM entered into after the issuance of the OMB guidance specified in subsection (a) terms mandating that the procured LLM complies with the Unbiased AI Principles, with provisions stipulating that decommissioning costs shall be charged to the vendor if the agency terminates the contract due to the vendor’s noncompliance after a reasonable period to remedy;
(ii) To the extent feasible and consistent with contract terms, amend existing contracts for LLMs to integrate the terms outlined in subsection (b)(i);
(iii) Within 90 days of the OMB guidance issued under subsection (a), establish procedures to ensure that LLMs procured by the agency align with the Unbiased AI Principles.
Sec. 5. General Provisions.
(a) Nothing in this order shall be interpreted to limit or otherwise affect:
(i) The authority vested by law in an executive department or agency, or its head; or
(ii) The functions of the Director of the Office of Management and Budget concerning budgetary, administrative, or legislative proposals.
(b) This order will be implemented in accordance with applicable law and subject to the availability of appropriations.
(c) This order is not intended to, and does not create any right or benefit, substantive or procedural, enforceable at law or in equity by any party against the United States, its departments, agencies, or entities, its officers, employees, or agents, or any other individual.
(d) The costs associated with the publication of this order shall be borne by the General Services Administration.
DONALD J. TRUMP
THE WHITE HOUSE,
July 23, 2025.
