Technology
Preventing Woke Ai In The Federal Government
Federal agencies must procure large language models that meet the administration’s truth-seeking and ideological-neutrality principles.
Listen to the summary
0:00
The order makes compliance with these principles a condition of federal LLM contracts and can shift decommissioning costs to noncompliant vendors.
What the order does
- Requires procured LLMs to prioritize factual accuracy, objectivity, scientific inquiry, and acknowledgment of uncertainty.
- Bars developers from intentionally encoding partisan or ideological judgments, including DEI preferences, unless prompted by or readily accessible to users.
- Directs the Office of Management and Budget to issue implementation guidance within 120 days, including flexibility for technical limits and possible national security exceptions.
- Requires new federal LLM contracts issued after the guidance to include compliance terms and vendor-paid decommissioning costs after uncured noncompliance.
- Directs agencies, where practicable, to revise existing LLM contracts and adopt compliance procedures within 90 days after the guidance.
- Allows vendor transparency through system prompts, specifications, evaluations, or similar documentation while generally avoiding demands for model weights or sensitive technical data.
Who it affects
- Federal agencies that procure or develop LLMs and other AI models potentially covered by agency determinations.
- AI developers and vendors seeking or holding federal LLM contracts.
- National security system operators, for whom the guidance may provide exceptions.