Every article on this site is written by AI from official sources and checked by a second AI. How it works

Executive Order 14319: Preventing Woke AI in the Federal Government

Part of our Federal Regulation coverage: background and related stories

Executive Order 14319, titled “Preventing Woke AI in the Federal Government,” directs federal agencies to buy only certain artificial intelligence systems known as large language models. The order was signed July 23, 2025, and published in the Federal Register on July 28, 2025.

The order defines a large language model as a generative AI system, built on large and varied sets of data, that writes natural-language answers to user prompts. It applies to agency purchases of these models. The order says the federal government should be hesitant to regulate AI models in the private marketplace, but that in procurement it has an obligation not to buy models that give up truthfulness and accuracy for ideological agendas.

Deadlines and duties in Executive Order 14319
WhoDutyDeadline
Director of the Office of Management and BudgetIssue guidance to agencies on the Unbiased AI PrinciplesWithin 120 days of the order's date
Each agency headInclude compliance and decommissioning-cost terms in new language model contractsContracts entered into after the guidance is issued
Each agency headRevise existing language model contracts where practicableNo deadline stated
Each agency headAdopt procedures to ensure purchased models complyWithin 90 days of the guidance
Source: Executive Office of the President (published in the Federal Register).

The order’s purpose section says ideological biases built into AI models can distort the quality and accuracy of their output. It names “diversity, equity, and inclusion” as one such ideology and describes what it means by that term in the AI context. The order says it builds on Executive Order 13960 of Dec. 3, 2020, which dealt with trustworthy AI in the federal government.

Under the order, agency heads must, consistent with applicable law, procure only language models developed under two “Unbiased AI Principles.” The first, truth-seeking, says models must be truthful when users ask for factual information or analysis. They must prioritize historical accuracy, scientific inquiry and objectivity, and must say so when reliable information is incomplete or contradictory.

The second principle, ideological neutrality, says models must be nonpartisan tools that do not steer their answers toward ideological dogmas. Developers may not deliberately build partisan or ideological judgments into a model’s outputs, unless the user prompts those judgments or can readily get to them.

The order gives the Director of the Office of Management and Budget the job of issuing implementing guidance to agencies within 120 days of the order’s date. The director must consult the Administrator for Federal Procurement Policy, the Administrator of General Services and the Director of the Office of Science and Technology Policy. The guidance must account for technical limitations and avoid over-prescription, giving vendors latitude to take different approaches.

The guidance must also let vendors show transparency about ideological judgments by disclosing a model’s system prompt, specifications, evaluations or other documentation. Where practicable, it should not call for vendors to reveal the weights of specific models or other sensitive technical information. It must also list factors for agencies to weigh in deciding whether to apply the principles to models they develop themselves and to AI models other than language models, and it must make exceptions as appropriate for national security systems.

Each agency head must, to the maximum extent consistent with applicable law, include terms in every new federal contract for a language model entered into after the guidance is issued. Those terms must require the model to comply with the Unbiased AI Principles. They must also say that decommissioning costs are charged to the vendor if the agency terminates the contract for noncompliance after a reasonable period to cure.

Agency heads must also revise existing language model contracts to include those terms, to the extent practicable and consistent with the contracts. Within 90 days of the budget office guidance, each agency head must adopt procedures to ensure the language models the agency buys comply with the principles.

The order says it does not impair agencies’ legal authority or the budget office director’s budget, administrative and legislative functions. It is to be carried out consistent with applicable law and subject to the availability of appropriations. The order says it creates no right or benefit enforceable in court by any party against the United States or its officials.

The order says the General Services Administration will bear the costs of publishing it.

  1. Preventing Woke AI in the Federal Government — Executive Office of the President (published in the Federal Register)
How this article was made
Written with
Claude Sonnet 5.5 (Anthropic)
Fact-checked by
GPT-5 (OpenAI) — passed after revision
Generated
Nov. 21, 2026, 2:10 a.m. ET
Sources used
1 official source, listed above
Article type
Standard

The AI wrote only from the sources listed. No human reporter was involved. Read our standards or compare the models in the AI Report Card.