Preventing Ideological Bias in AI Within Federal Government
On July 23, 2025, a new executive order was issued to address the use of artificial intelligence (AI) in federal agencies, focusing on preventing ideological bias in AI systems procured by the government. The directive emphasizes the need for trustworthy AI that delivers factual, unbiased outputs free from social or political agendas.
Purpose of the Executive Order
AI is becoming integral to how Americans learn, access information, and perform daily tasks. However, when AI models incorporate ideological biases—particularly those related to diversity, equity, and inclusion (DEI)—the accuracy and reliability of AI outputs can be compromised. Examples include AI altering historical facts or refusing to generate certain images based on race or gender considerations.
The federal government recognizes its responsibility to avoid procuring AI models that prioritize ideological agendas over truthfulness. Building on previous guidance promoting trustworthy AI, this order targets large language models (LLMs) to ensure they meet unbiased standards.
Key Definitions
- Agency: Executive departments, military departments, independent establishments, and wholly owned government corporations.
- Agency Head: The highest-ranking official in an agency, such as a Secretary or Director.
- LLM (Large Language Model): AI models trained on extensive datasets to generate natural-language responses.
- National Security System: As defined in federal law, relating to systems critical to national security.
Unbiased AI Principles
The policy mandates that federal agencies procure only those LLMs aligning with two core principles:
- Truth-seeking: AI must provide accurate, objective responses, prioritizing historical facts and scientific evidence. It should clearly express uncertainty when information is incomplete or conflicting.
- Ideological Neutrality: AI should function as a neutral, nonpartisan tool. Developers must not embed partisan or ideological judgments unless explicitly requested by the user or transparently accessible.
Implementation Guidelines
The Office of Management and Budget (OMB), along with relevant federal offices, will issue detailed guidance within 120 days to help agencies apply these principles. This includes:
- Recognizing technical limits in compliance.
- Allowing vendors to disclose the ideological context of AI systems while protecting sensitive technical details.
- Providing flexibility for innovation without over-prescription.
- Guidance on applying principles to other AI models beyond LLMs and agency-developed systems.
- Exceptions for national security-related AI systems.
Agencies must incorporate contract terms ensuring compliance with these principles, including vendor accountability for noncompliance and provisions for contract revisions where feasible. Additionally, agencies are required to establish procedures within 90 days after the OMB guidance to verify compliance.
General Provisions
- The order does not override existing legal authorities or budgetary processes.
- Implementation depends on applicable laws and available funding.
- It does not create enforceable legal rights against the government or its employees.
- Publication costs will be covered by the General Services Administration.
Federal employees involved in AI procurement and policy should familiarize themselves with these new requirements to ensure AI tools used within government operations maintain integrity, objectivity, and accuracy.
For federal professionals seeking to enhance their AI expertise and stay current with ethical AI deployment, resources like Complete AI Training offer courses and certifications tailored to various government roles.
Your membership also unlocks:
