The Defense Department has issued new guidance for contracting and program personnel designed to ensure that the artificial intelligence systems the Pentagon buys adhere to the Trump administration's "truth-seeking" and "ideological neutrality" procurement principles, Federal News Network reported Monday.
The directive, issued by John Tenaglia, the Pentagon's principal director for defense pricing, contracting and acquisition policy, is the department's most concrete step yet to implement President Donald Trump's agenda to "prevent woke AI" in the federal government. For more context on this story, see our ongoing more AI stories.
From Executive Order to Contract Language
The guidance flows directly from the executive order Trump signed in July, which requires federal agencies to ensure the large language models they purchase are "neutral, nonpartisan tools that do not manipulate responses in favor of ideological dogmas such as [diversity, equity, and inclusion]." In that order, Trump declared that federal agencies "have the obligation not to procure models that sacrifice truthfulness and accuracy to ideological agendas."
The Office of Management and Budget later issued a memo addressing questions that emerged after the executive order was released. The Pentagon's new document translates those high-level mandates into the machinery of defense acquisition: templates that walk officials through different stages of buying AI systems and verifying contractor compliance with the department's "unbiased AI" principles.
Specifically, the templates include a memo to help agencies select appropriate transparency and documentation requirements for a procurement, standardized solicitation language telling companies what documentation they must provide to demonstrate compliance, and formal contract language that can be incorporated into a statement of work or performance work statement, according to Federal News Network.
Guidance, Not a Mandate — For Now
Tenaglia was careful to frame the document's legal weight. While compliance with the underlying executive order and the OMB memo is mandatory, the new document is intended "as guidance only," providing "a standardized approach to documenting, soliciting and enforcing the unbiased AI principles," he wrote.
"Recognizing that AI procurements vary widely in complexity and risk, DoD components and requiring activities have the flexibility to revise, tailor or supplement this language as necessary to fit the specific needs and context of their individual requirements," Tenaglia wrote in the memo.
That flexibility may prove consequential. Measuring whether an AI model is "unbiased" — and documenting that determination for a contract file — is a far messier task than checking a cybersecurity certification, and vendors are likely to push back on documentation demands that touch on model design.
Part of a Bigger Push to Commercialize Defense AI
The guidance arrives as Defense Secretary Pete Hegseth continues his push to bring commercial AI into the department. Earlier this year, the Pentagon issued an AI strategy aimed at turning the department into an "AI-first warfighting force," with Hegseth saying the department will invest heavily in AI compute infrastructure, from data centers to systems at the tactical edge.
Lawmakers are moving in parallel. Rep. Jimmy Patronis (R-Fla.) is looking to codify Trump's executive order into law. "The president and his team have delivered a smart executive order that instructs federal agencies to only leverage AI tools that speak the truth," Patronis said, arguing there are "examples-after-examples of AI tools going off the rails because the foundational programming tilts politically left."
What It Means for AI Vendors
For AI companies — including the frontier labs now racing toward public offerings — the guidance signals that ideological-neutrality documentation is becoming a de facto requirement for the U.S. defense market, one of the fastest-growing customer segments for commercial AI. Vendors that can produce audit trails for their models' training and behavior may find the door open; those that cannot may find federal procurement increasingly closed.
The practical effect is that "unbiased AI" is graduating from political rhetoric to a written requirement in the world's largest technology buyer's contracts — and every other agency is watching how the Pentagon's templates perform.
What Vendors Will Actually Have to Produce
Under the templates, the burden of proof shifts toward the vendor. Companies bidding on AI-heavy contracts would need to supply documentation demonstrating that their models comply with the "unbiased AI" principles — effectively creating a new category of acquisition paperwork alongside security accreditations and supply-chain attestations. Contracting officers, in turn, get standardized language they can drop into solicitations without drafting from scratch, which procurement experts say is usually what turns policy into practice.
The guidance also arrives as the Pentagon normalizes commercial chatbots for its workforce, having rolled out government-tuned versions of commercial AI assistants for defense personnel. Embedding ideological-neutrality language in those contracts extends a political standard into systems used for everything from logistics planning to routine administrative work.
A Test Case for the Rest of Government
Other agencies now face a choice: adopt the Pentagon's templates wholesale, write their own, or wait for Congress to act. Patronis's push to codify the executive order would resolve that ambiguity by statute — and would make neutrality requirements far harder for a future administration to reverse.
Critics of the order argue that measuring ideological neutrality in language models is scientifically fraught and could chill legitimate model behavior, including accurate descriptions of contested topics. Supporters counter that procurement standards already routinely impose content requirements on vendors and that model developers remain free to sell elsewhere.
Either way, the direction of travel is clear. The government's largest technology buyer has put "unbiased AI" in writing, and defense contractors hoping to sell AI systems to the Pentagon now have a new checklist to satisfy — one that starts with documentation and ends, potentially, in contract disputes if a deployed model fails the neutrality test.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
