Companies that sell large language models to the Pentagon now face standardized contract language requiring them to document that their models meet the Trump administration's "truth-seeking" and "ideological neutrality" principles, and military buyers have been told that compliance is mandatory. The Defense Department's acquisition policy office issued templates that let contracting officers solicit, document and enforce those requirements, DefenseScoop reported on Oct. 23. A memorandum dated Oct. 20 from John Tenaglia, principal director for Defense Pricing, Contracting, and Acquisition Policy, says the guidance is "in direct support" of President Trump's executive order on "Preventing Woke AI in the Federal Government" and an Office of Management and Budget memo on "Increasing Public Trust in Artificial Intelligence Through Unbiased AI Principles."
The memo went to acquisition leaders at the Departments of the Army, Navy and Air Force, U.S. Special Operations Command, U.S. Cyber Command, U.S. Transportation Command, and DOD agencies and field activities. Tenaglia noted that compliance with the White House directives is mandatory, according to DefenseScoop.
Three Templates for Buying 'Unbiased' LLMs
The guidance arrives as three attachments aimed at the people who write and award AI contracts. The first is a memorandum for the requiring activity, the office that needs the capability, to select the transparency requirements appropriate to the procurement. The second is standardized solicitation language that tells vendors what documentation they must provide to demonstrate compliance. The third is formal contract language that can go into the Statement of Work or Performance Work Statement in solicitations and final contracts.
The issuance "provides contracting and program personnel with adaptable templates and standardized language to document, solicit, and enforce these Unbiased AI Principles in our procurements. It is designed to be flexible, allowing DoW Components to tailor the language to fit the specific context and risk level of their individual AI acquisitions," Tenaglia wrote, as quoted by DefenseScoop. "DoW" refers to the Department of War, the administration's preferred name for the Defense Department.
The memo also suggests that contract language define key terms, clearly communicate submission requirements to potential vendors, and make the provision of compliance documentation a formal part of the proposal process. That last point matters for bidders: if the documentation is part of the proposal, a missing or thin submission becomes an evaluation problem rather than a post-award paperwork item.
Why a Reseller's Disclosure Burden Differs From a Model Developer's
When the Pentagon buys an LLM, it must obtain enough information from the contractor to determine whether the technology complies with the principles in Trump's order, according to an attachment to Tenaglia's memo. The department acknowledges that not every vendor can answer the same questions. "The amount and type of information available will vary depending on the contractor's role within the software supply chain and its relationship with the LLM developer itself, with more information generally being available from sources closer to the original LLM developer," the attachment states, per DefenseScoop.
OMB's underlying memo anticipates the same split. Its appendix addresses vendors that are not direct developers of the LLM and points to the layers they may control instead, such as classifiers, system prompts, fine-tuning, and content moderation and filters. The templates give contracting officers room to ask a model developer for more than they ask of a downstream integrator.
What OMB M-26-04 Already Requires
The Pentagon guidance implements OMB Memorandum M-26-04, signed by OMB Director Russell Vought on Dec. 11, 2025. It defines the two principles from Executive Order 14319. Truth-seeking LLMs "shall be truthful in responding to user prompts seeking factual information or analysis." Under ideological neutrality, LLMs "shall be neutral, nonpartisan tools that do not manipulate responses in favor of ideological dogmas."
The OMB memo applies to any LLM procured by an agency. It does not apply to national security systems as defined in 44 U.S.C. 3552(b)(6), though OMB encourages applying it to those systems where practicable. Any solicitation or order for an LLM issued after the memo's date must include contractual requirements addressing compliance with the Unbiased AI Principles. For existing contracts, agencies should modify them to add the requirements, at the latest before exercising any option that extends the period of performance. Agencies had until March 11, 2026, to update their procurement policies, including processes for users to report outputs that violate the principles.
The enforcement hook sits in the memo's appendix. "Agencies should explicitly identify relevant requirements identified in (a) and (b) as material to eligibility and payment under the contract, to support termination of the contract for default, as directed by the E.O. in cases where the vendor refuses to take corrective action in identified cases of noncompliance," the memo states.
OMB also set a limit on how far documentation demands should reach. "Where practicable, agencies should avoid requirements that compel a vendor to disclose sensitive technical data, such as specific model weights," the memo says. The memo sunsets two years after issuance unless the OMB director provides otherwise.
How Much Room Do Components Have to Tailor the Language?
Tenaglia's memo describes the templates as a starting point, not a fixed clause. Components may tailor the language to the context and risk level of each AI acquisition.
That flexibility cuts both ways for industry. Vendors may see different versions of the same requirement across the Army, Navy, Air Force and the commands named in the memo, and proposal teams will need to read each solicitation's version closely rather than reuse a single compliance package.
What It Means for Contractors
Firms that sell or embed LLMs in defense work should expect Unbiased AI compliance documentation to appear as a proposal requirement, with the governing terms written into the Statement of Work or Performance Work Statement. Because OMB directs agencies to treat the requirements as material to eligibility and payment, a vendor that refuses corrective action after a finding of noncompliance faces possible termination for default.
Incumbents are exposed too. OMB's guidance points agencies to modify existing LLM contracts no later than the next option exercise that extends performance, so a contractor approaching an option year should prepare for a modification adding these clauses.
Integrators and resellers should map what they can actually document about the models they deliver, from system prompts to fine-tuning and filters, and line up information from their model suppliers. The Pentagon's own attachment concedes that more information sits closer to the original developer. OMB's caution against compelling disclosure of model weights gives vendors a basis to push back if a solicitation asks for that level of technical data.
