The federal government wins pre-release evaluation rights on Google, Microsoft, and xAI models

verified

Friday, May 8, 2026

The News

The Center for AI Standards and Innovation announced agreements with Google DeepMind, Microsoft, and xAI giving the U.S. government access to evaluate AI models before public release. Anthropic and OpenAI are not currently part of the program. The framework is the first formal pre-release model audit by federal authorities.

The Read

Three of the largest AI providers just signed up to add a federal evaluation step to their launch calendar; two of the largest did not. That's the headline: pre-release model audit is now an asymmetric competitive variable. For a buyer, the question is no longer "which model is best" — it's "which model has cleared which audit by what date," and which procurement workflows you can defend if a vendor's release ships with a federal evaluation hanging over it. Read it as a foreshadowing of how AI procurement starts looking more like medical-device or financial-product procurement: regulator-scoped, timeline-coupled, and increasingly bifurcated by which side of the federal review your vendor sits on.

Watch For

Draft a one-page memo for your product and legal leadership outlining what your team would do if your model provider had to delay a release for federal pre-evaluation — specifically, which features are gated by the upcoming model version, how you'd communicate slip to customers, and which hedge providers you'd line up. Circulate it before Q3 planning. The exercise is the deliverable; if you can't draft the memo in 20 minutes, you don't yet have a clear-enough dependency map on your model layer.

For Product