Preventing Woke AI in the Federal Government
Headline: Federal Agencies Required to Purchase Only Unbiased Large Language Models
What it does: Agencies must procure only LLMs that meet the Unbiased AI Principles and add contract terms enforcing compliance under OMB guidance.
- Vendors may have to disclose system prompts, specifications, or evaluations to federal buyers.
- Contracts can require vendors to pay decommissioning costs if they fail to comply.
- Agencies must update contracts and adopt compliance procedures within set timeframes.
Summary
This order requires federal agencies to buy only large language models (LLMs) that follow two "Unbiased AI Principles": truth-seeking and ideological neutrality. The Office of Management and Budget must issue guidance within 120 days to explain how agencies should apply those principles.
Agency leaders must add contract terms requiring compliance, try to update existing LLM contracts, and adopt procedures within 90 days after OMB guidance. Vendors may need to disclose system prompts or specifications and can face decommissioning costs for noncompliance; national security systems may be excepted.
Ask about this order
Ask questions about this executive order and its implications.
What agencies are affected by this order?
How does this order change existing policy?
What are the practical implications of this order?