United States of America: White House Office of Management and Budget adopted memorandum on increasing public trust in artificial intelligence through unbiased AI principles (M-26-04)

On 11 December 2025, the Director of the Office of Management and Budget issued a memorandum on increasing public trust in artificial intelligence by setting out unbiased AI principles. The memorandum was in line with Section 4 of Executive Order 14319 to guide agencies in implementing AI principles and to complement earlier OMB guidance on AI acquisition. It applies to executive and military departments, independent establishments, and wholly owned government corporations, covering any Large Language Model procured by an agency and encouraging similar practices for national security systems where suitable. Agencies are required to include contractual obligations ensuring compliance with the two “Unbiased AI Principles” in all new solicitations and orders for LLMs and other relevant AI models issued after the memorandum date. The two Unbiased AI Principles require, first, that truth-seeking LLMs provide truthful responses to user prompts seeking factual information or analysis, prioritising historical accuracy, scientific inquiry, and objectivity, and acknowledging uncertainty where information is incomplete or inconsistent. Second, LLMs must maintain ideological neutrality and operate as nonpartisan tools that do not shape responses to favour particular ideological positions. Developers are instructed not to intentionally encode partisan or ideological judgements into model outputs unless such content is expressly prompted by, or readily accessible to, the user. They are also expected to modify existing contracts for LLMs, where practicable, before extending performance. By 11 March 2026, agencies must update procurement policies to reflect these principles and establish mechanisms allowing users to report any violations. The memorandum sets out minimum transparency obligations for LLM vendors, including the provision of an Acceptable Use Policy, Model, System, or Data Cards, end-user resources, and a channel for feedback. Agencies should designate these obligations as material contract requirements to allow termination if a vendor fails to comply.