# Executive summary — GS Federal AI Procurement Safeguards Index

The Codex-reviewed index is **31.5 out of 100**, calculated as the equal-weight mean
of ten reviewed safeguard prevalence estimates across public federal AI solicitation
packages from January 2024 through August 2026.

- Highest estimated safeguard: **Performance Metrics** —
  59.7%.
- Lowest estimated safeguard: **High Impact Disclosure** —
  3.3%.
- Candidate notices: **1,532**; scoreable public text:
  **1,076**.
- Classifier precision: **95.6%**; recall:
  **96.3%**.

These results describe public procurement language, not unobservable internal agency practices.
The pre/post M-25-22 comparison is descriptive and noncausal.