EEO Compliance With AI Recruiting Tools: A Team Playbook

Recruitment teams keep AI hiring tools EEO compliant by treating them like any other selection procedure: use job-related criteria, test outcomes for adverse impact by group, offer accommodations and alternatives, keep records, and hold vendors to clear contract terms. Using AI does not shift legal responsibility. The employer remains accountable for discriminatory outcomes.

Key takeaways

  • Federal anti-discrimination laws apply to AI tools the same way they apply to tests and interviews.
  • The four-fifths rule is a starting screen for adverse impact, not the final answer.
  • Candidates with disabilities need a clear path to request accommodation or an alternative process.
  • Some jurisdictions, such as New York City, add bias audit and notice requirements.

Which EEO rules apply to AI in hiring?

The EEOC enforces federal laws that bar discrimination based on race, color, religion, sex (including pregnancy, sexual orientation and gender identity), national origin, age (40 or older), disability and genetic information. Those laws cover the outcome of a hiring process, whether a human or software made the decision. An AI tool that screens out a protected group at a higher rate, or that a candidate with a disability cannot use, can create liability for the employer.

The Uniform Guidelines on Employee Selection Procedures, published in the Code of Federal Regulations, explain how federal agencies evaluate selection procedures. They apply to scored or ranked screening steps, including automated ones.

How do you check an AI tool for adverse impact?

Under the Uniform Guidelines, a selection rate for any race, sex or ethnic group that is less than four-fifths (80 percent) of the rate for the highest group is generally treated as evidence of adverse impact. The guidelines also note that smaller differences can matter if they are statistically and practically significant, and that small samples may need a longer review period.

A practical cadence:

  1. Record applicants and outcomes at each AI-assisted step, such as screened in, advanced to interview and hired.
  2. Calculate selection rates by group for each step and for the process overall.
  3. Flag any ratio under four-fifths and run a statistical significance test with expert help.
  4. Investigate which criteria drive the gap and whether they are job-related.
  5. Adjust or remove the criteria, and document what changed and why.

Our step-by-step guide on self-auditing your ATS for disparate impact walks through the calculations.

Compliance checklist for recruiting teams

AreaWhat to do
Job criteriaBase screening on documented, job-related requirements; remove proxies such as graduation year or gaps
Human reviewHave a trained recruiter review rejections and edge cases rather than auto-rejecting
AccommodationTell candidates how to request accommodation or an alternative format before they start
NoticesDisclose AI use where required and where it builds trust
MonitoringRun adverse impact checks on a set schedule and after configuration changes
RecordsKeep applicant data, scores and decisions as required by recordkeeping rules
VendorsGet validation evidence, audit support and data handling terms in the contract

Design choices in the tool itself matter too. Hiding names and photos during early review is one option, discussed in our article on blind candidate ranking, and our guide to fair hiring guardrails for AI resume screening covers configuration settings.

What about state and local AI hiring laws?

Several jurisdictions have added their own rules. New York City's Local Law 144 is the best-known example: employers and employment agencies may use an automated employment decision tool only if it has had a bias audit within one year, a summary of the results is posted publicly, and candidates receive notice. The city's Department of Consumer and Worker Protection began enforcing the law in July 2023 and states that notice must be given 10 business days before use. Other states and cities have passed or proposed related rules, so check the locations where you hire and review them at least annually.

Questions to ask any vendor

  • How was the tool validated for the jobs we will use it for?
  • Can you provide adverse impact results by group, and will you support our own audits?
  • Which candidate data does the model use, and can we turn off specific factors?
  • How do candidates request an accommodation or alternative process?
  • What records do you keep, for how long, and can we export them?

Frequently Asked Questions

Is the employer or the AI vendor responsible for discrimination?

The employer remains responsible for its hiring decisions under federal anti-discrimination law, even when a vendor's tool produces the result. Vendors may share liability in some situations, and contracts can allocate costs, but recruiting teams should not assume the vendor has handled compliance. Ask for evidence, run your own checks and document them.

Does passing the four-fifths rule mean a tool is compliant?

Not on its own. The four-fifths rule is a rule of thumb for spotting adverse impact. Smaller differences can still be significant, and a tool can raise other issues, such as screening out people with disabilities or relying on criteria that are not job-related. Use it as a first screen alongside statistical tests and validation evidence.

How often should we audit AI hiring tools?

At least annually, and whenever the tool, its configuration or the jobs it screens for change significantly. High-volume roles may justify quarterly checks. In New York City, a bias audit within one year of use is legally required. Keep each audit, its findings and any fixes on file.

Next steps

Inventory every AI-assisted step in your hiring process and name an owner for each. Pull the last six months of outcome data, run a four-fifths check by group and review your accommodation language. Then schedule vendor conversations using the questions above and add audit support to renewal terms.

Last reviewed: October 2026.

Sources

Comments

Popular Posts

Why Workday Onboarding Breaks Down for Frontline Employees

How to Improve the Customer Experience (CX)

How Do I Log In and Sign In to Workday HCM

UKG Union Contract and Overtime Rules for Multi-Site Food Plants

Infor HCM Software Engineer Jobs: Salary Factors and Skills

ERP Solution Guide: How to Choose the Best ERP for You

How to Search in Workday: Find Reqs, Reports, People and Tasks

The End of Remote Work: Why Companies Want You Back

New Apple Watch Health Features Will Be Available This Year, but Blood Pressure and Blood Sugar Sensors Will Not Be Available Until Next Year

The Future of Employee Healthcare Key Concerns and Strategies for 2024