Executive Order 14319: Preventing Woke AI in the Federal Government
Issued by
White House, Executive Office of the President
- September 30, 2026 · Correction — Rewrote the entry: signed 23 July 2025 as binding executive action, scoped to LLM procurement, implemented by OMB M-26-04, with its actual exceptions; it revokes nothing (fact-check finding). (Cody Maxwell)
- October 1, 2026 · Correction — Typed as an executive order, a new directory category. (Cody Maxwell)
- October 1, 2026 · Correction — Rewrote the practical steps, newsletter hook, search description, and audience fields to match the entry's corrected content. (Cody Maxwell)
Executive Order 14319, signed on 23 July 2025, directs federal agencies to buy only large language models that follow two Unbiased AI Principles: truth-seeking and ideological neutrality. OMB implemented it in Memorandum M-26-04 on 11 December 2025.
Applies To
Overview
President Trump signed Executive Order 14319 on 23 July 2025, and it was published in the Federal Register on 28 July 2025. It took effect on signing. The order directs agency heads to procure only large language models developed in line with two Unbiased AI Principles. Truth-seeking means models answer factual questions accurately and acknowledge uncertainty. Ideological neutrality means models do not manipulate answers toward ideological positions. Federal LLM contracts must include compliance terms, with vendors bearing decommissioning costs if a contract is ended for noncompliance. OMB issued the implementing guidance, Memorandum M-26-04, on 11 December 2025. It requires these terms in any LLM solicitation or order issued after that date, and required agencies to update procurement policies by 11 March 2026. National security systems are excepted, as are LLMs under free open-source licenses and incidental contractor use. The order builds on Executive Order 13960 on trustworthy AI in government and revokes nothing. Biden's EO 14110 had already been revoked by EO 14148 in January 2025.
Key Requirements
- •Agencies may procure only large language models that follow the truth-seeking and ideological neutrality principles.
- •Federal LLM contracts must include compliance terms, with vendors paying decommissioning costs if terminated for noncompliance.
- •OMB M-26-04 applies these terms to LLM solicitations and orders issued after 11 December 2025.
- •Agencies had to update procurement policies by 11 March 2026, including a way for users to report violating outputs.
- •National security systems, free open-source LLMs, and incidental contractor use are excepted.
- •Agencies may extend the principles to other AI models at their discretion.
What Your Organization Must Do
- →Map which of your federal contracts supply large language models and could face the new procurement terms.
- →Prepare model cards and supporting documents showing accurate answers to factual questions and acknowledgment of uncertainty.
- →Review contract templates for the compliance terms, including vendor-borne decommissioning costs on termination for noncompliance.
- →Check whether your offering falls within the exceptions for national security systems, free open-source models or incidental contractor use.
- →Add the terms to existing federal LLM contracts where practicable, and at the latest before any extension.
- →Set up an internal channel for agency users to report outputs that appear to breach the principles.
Playbook Guidance
Step-by-step implementation guidance for compliance teams.
Frequently Asked Questions
- Which federal agencies are required to comply with EO 14319?
- Executive departments, military departments, independent establishments, and wholly owned government corporations, for their LLM purchases. National security systems, free open-source LLMs, and incidental contractor use are excepted.
- What is the compliance deadline for EO 14319?
- Contract terms apply to LLM solicitations and orders issued after 11 December 2025, when OMB issued M-26-04. Agencies had to update procurement policies by 11 March 2026.
- What does EO 14319 mean for AI vendors with existing federal contracts?
- Agencies should add the terms to existing LLM contracts where practicable, at the latest before extending them. Vendors must supply documents such as model cards and bear decommissioning costs if terminated for noncompliance.
- What are the penalties under EO 14319?
- The order sets no fines. The consequence for vendors is contractual: termination for noncompliance, with decommissioning costs. Agency users report problem outputs through internal channels, not to OMB.
- How does EO 14319 relate to earlier Biden-era AI executive orders?
- EO 14319 revokes nothing; it builds on Executive Order 13960 on trustworthy AI in government. Biden's EO 14110 was revoked separately by EO 14148 on 20 January 2025.
- What specific AI output characteristics does EO 14319 address in federal procurement?
- Models must answer factual questions truthfully and acknowledge uncertainty, and must not manipulate answers toward ideological positions. Neither the order nor M-26-04 sets a numeric test for neutrality.