Source: The White House

The White House published an executive order aimed at preventing ideological bias, particularly related to diversity, equity, and inclusion (DEI), in AI systems procured by the federal government. The order establishes two core principles for large language models (LLMs) used by federal agencies: truth-seeking and ideological neutrality. It mandates that LLMs prioritize historical accuracy, scientific objectivity, and avoid partisan or ideological manipulation unless explicitly prompted.

The Office of Management and Budget (OMB) will issue guidance within 120 days to implement these principles, including contract terms requiring vendor compliance and provisions for decommissioning non-compliant AI systems. The order excludes national security systems from some requirements and emphasizes transparency while avoiding over-prescription. This directive reflects the administration’s stance against what it terms “woke AI” in federal procurement.