President Donald Trump took decisive action against artificial intelligence models that prioritize diversity and inclusion over factual accuracy in federal operations.
According to The White House, the executive order signed on July 23, 2025, mandates that federal agencies can only procure AI models that adhere to principles of truth-seeking and ideological neutrality. The order specifically targets large language models (LLMs) that incorporate concepts like critical race theory, transgenderism, and systemic racism.
The executive order comes in response to several controversial incidents involving AI models that altered historical images and facts to align with diversity, equity, and inclusion (DEI) principles. These incidents included changing the race of historical figures like the Pope and Founding Fathers, refusing to generate images celebrating specific racial groups, and promoting ideological positions over factual accuracy.
Federal contractors developing AI models must now prove their systems prioritize historical accuracy and scientific objectivity over ideological considerations. The Office of Management and Budget (OMB) will issue detailed implementation guidance within 120 days of the order.
Agency heads are required to include specific contract terms ensuring AI models comply with the new "Unbiased AI Principles." These principles emphasize truthfulness in responding to user prompts and neutrality in handling ideological or partisan content.
Vendors found non-compliant with these requirements will face potential contract termination and bear the costs of decommissioning their AI systems. The order allows for technical limitations and aims to avoid over-prescription while promoting innovation.
The executive order includes provisions for national security systems, with the OMB guidance expected to outline specific exceptions. This recognizes the unique requirements of AI applications in sensitive government operations.
Federal agencies have 90 days after receiving OMB guidance to establish procedures ensuring their procured AI models comply with the new principles. The implementation timeline reflects the urgency of addressing AI bias while allowing reasonable adaptation periods.
Existing contracts for AI systems will be revised where practicable to include the new requirements, showing the administration's commitment to comprehensive reform of federal AI procurement.
The order establishes clear definitions for affected agencies, including executive departments, military departments, and independent establishments. This broad scope ensures consistent application across the federal government.
Agency heads maintain significant authority in determining how to apply these principles to AI models developed internally. The guidance will provide factors for consideration while allowing flexibility in implementation approaches.
The administration emphasizes that the order doesn't create new legal rights or benefits enforceable against the United States. Instead, it focuses on internal government procurement and operational standards.
The executive order builds upon previous initiatives, particularly Executive Order 13960 from December 2020, which promoted trustworthy AI in federal government operations. This new measure specifically addresses large language models and their potential ideological biases.
Federal agencies must now evaluate AI systems based on their ability to acknowledge uncertainty when information is incomplete or contradictory. This requirement aims to prevent AI models from making definitive statements without sufficient factual basis.
These procurement changes reflect broader concerns about AI's role in shaping public discourse and decision-making, particularly in government operations where accuracy and neutrality are paramount.
