Connect with us

Business

US to impose unbiased AI requirements on federal contractors

Published on

Credit: Igor Omilaev

The White House has released new government-wide guidance directing federal agencies to ensure that all large language models procured by the U.S. government comply with newly established “Unbiased AI Principles,” according to a memorandum from the Office of Management and Budget (OMB).

The memorandum, issued on Thursday, outlines implementation requirements stemming from President Trump’s July executive order on preventing “woke AI” in federal systems.

The guidance mandates that new federal contracts for large language models include provisions ensuring compliance with two principles: “truth-seeking” and “ideological neutrality,” according to the document.

According to the memorandum, “truth-seeking” requires that AI systems provide factual, objective responses while acknowledging uncertainty where information is incomplete or contradictory.

“Ideological neutrality” requires that systems avoid partisan or ideological manipulation and prohibits developers from intentionally encoding ideological judgments unless those judgments are prompted by the end user.

The policy applies to all LLMs procured by executive agencies and encourages, but does not require, application to national security systems.

The guidance instructs agencies to update procurement policies by March 11, 2026, to ensure compliance, and requires federal contracts to include mechanisms for reporting AI outputs that violate the principles.

Agencies are also directed to collect documentation from AI vendors, including acceptable-use policies, model or system cards, evaluation materials, and feedback mechanisms, to assess whether a model meets the standards established by the executive order.

In addition to new contracts, agencies are encouraged to modify existing AI contracts before exercising renewal options to bring them into compliance.

The memorandum said the requirements are intended to strengthen public trust in federal uses of artificial intelligence by ensuring that models used by the government adhere to consistent, nonpartisan standards.

Most Viewed