OMB M-26-04: Right-Size LLM Transparency Evidence
Explore how OMB Memorandum M-26-04 sets transparency requirements for acquiring large language models (LLMs) in federal agencies. Learn to structure procurement packages with enforceable terms, including minimal and enhanced transparency evidence, to support monitoring, compliance with unbiased AI principles, and manage vendor updates. Gain skills to demand documentation like acceptable use policies and feedback mechanisms, ensuring your contract supports oversight and risk management in AI deployments.
A procurement questionnaire can look complete while still leaving the team unable to govern the LLM they are buying. The common failure is a form that asks for price, uptime, and a baseline security posture, but never asks how the model will change, how users will be constrained, or how the agency will learn about failures. That gap does not show up on day one, because the first demo works and everyone thinks the hard part is picking a vendor. The work that actually decides whether the buy is defensible is whether the contract can support the agency’s existing ...