OMB M-25-21 · 29 CFR part 1607

AI Hiring Bias Audit for Government

An AI hiring bias audit for government measures whether an AI screening or ranking tool selects groups at different rates (the four-fifths rule and significance tests under 29 CFR part 1607) and documents the pre-deployment testing and monitoring OMB M-25-21 requires for high-impact AI. We do not audit tools we built.

What the Rules Require

  • OMB M-25-21 (April 3, 2025): pre-deployment testing, an impact assessment and ongoing monitoring for high-impact AI, which includes AI used in employment decisions.
  • 29 CFR 1607.4(D): the four-fifths rule as the general measure of adverse impact.
  • OPM's memo on AI in federal hiring applies M-25-21 to hiring uses.

Our Method

  1. Define the decision points the tool affects.
  2. Collect outcomes by group, protecting privacy (groups under five suppressed).
  3. Impact ratios, significance tests and intersectional checks.
  4. Findings, mitigations and a monitoring schedule.

Sample Deliverable

The audit's results table:

  • 1. Tool and use
  • 2. Data
  • 3. Results
  • 4. Limitations
  • 5. Recommendations
AI screen pass rates. SAMPLE: fictional data, not from any agency or client.
GroupScreenedPassedImpact ratioFisher's exact p
Group A5002501.00—
Group B3001350.900.18
Group C120420.70< 0.01

Work With Us

Need an I-O psychology subcontractor for your next human-capital bid? Send the vehicle and what you need. No call required.

Send a teaming request Federal capabilities

Related

Free tools and research

Questions

Can we try the math ourselves first?

Yes. The free four-fifths rule calculator runs the same impact-ratio check in your browser.

Sources (Checked November 18, 2026)

General information, not legal advice.