AI in Hiring, Performance & Termination
Hiring and performance tools can affect employment decisions even when a manager makes the final choice. These answers explain what to review, what to document, and where human judgment is needed.
Questions and Answers
Can employers use AI to screen job applicants?
Yes, but using software does not remove the employer’s obligations under employment discrimination law. Review what the tool measures, whether that measure relates to the job, and whether applicants can complete the process with an appropriate accommodation.
Before launch, identify who can challenge or override a score. Keep the tool’s settings, vendor explanations, and the reasons for hiring decisions. A vendor’s claim that a product is compliant is not a substitute for reviewing how your organization uses it.
Sources: EEOC: Employment Tests and Selection Procedures; EEOC: Disability Discrimination.
Do changes to federal AI guidance remove discrimination obligations?
No. An agency can withdraw guidance or change enforcement priorities without repealing the underlying statutes. Employers should distinguish an agency’s current guidance from the laws and court decisions that apply to a hiring practice.
When updating a policy, check the status of the cited source and the law itself. Do not treat an archived document as current guidance, or assume that removing a web page makes the underlying practice lawful.
What should Texas employers know about the state's AI law?
Texas HB 149 took effect on January 1, 2026. It includes restrictions on developing or deploying AI with the intent to discriminate unlawfully. Its intent standard is not a replacement for federal employment discrimination rules, which can also reach practices with unlawful discriminatory effects.
Whether a particular tool or use falls within the Texas law requires a separate review. Employers should examine the product’s actual function, the information it uses, and the decisions it influences instead of relying only on its AI label.
Sources: Texas Legislature: HB 149 Enrolled Text; EEOC: Employment Tests and Selection Procedures.
What is changing for automated employment decisions in Colorado?
As of October 21, 2026, Colorado’s revised Automated Decision-Making Technology Act is scheduled to take effect January 1, 2027, and implementing rules are still being developed. The law addresses covered technology used to materially influence consequential decisions, including employment decisions.
Employers should inventory relevant tools, identify affected applicants and employees, and review notice, explanation, data-correction, human-review, and recordkeeping requirements against the final law and rules. Not every software feature has the same obligations. Confirm the current requirements before implementation.
Sources: Colorado Attorney General: Automated Decision-Making Technology.
Can a Wyoming employer assume that AI hiring tools need no review?
No. Federal employment laws and Wyoming’s Fair Employment Practices Act can apply to hiring decisions regardless of the technology involved. The absence of a rule aimed at a particular product does not establish that the employer’s use is lawful.
Location also matters. A Wyoming business recruiting or employing people in another state should assess that state’s requirements rather than relying only on the location of its headquarters.
Sources: Wyoming Legislature: Labor and Employment Statutes, Title 27; EEOC: Employment Tests and Selection Procedures.
Who is responsible if a vendor's tool produces discriminatory results?
An employer should not assume that purchasing a third-party tool transfers its employment-law responsibilities to the vendor. Potential liability depends on the employer’s actions, the vendor’s role, and the law supporting the claim.
Review the agreement’s cooperation, access-to-records, indemnification, and liability provisions. Confirm who will explain a disputed result and provide the records needed to investigate it. Contract protections may allocate costs between the parties, but they do not prevent an applicant or employee from bringing a claim.
Sources: EEOC: Employment Tests and Selection Procedures; NIST: AI Risk Management Framework.
Does passing the four-fifths test prove that a hiring tool is fair?
No. The four-fifths rule is a screening measure used when reviewing selection rates; it is not a guarantee that a process is lawful. A rate below 80 percent of the highest group’s rate can indicate adverse impact, but sample size, statistical significance, and other evidence matter.
Review the full selection process and the job-related reasons for its criteria. A favorable ratio does not resolve disability accommodation issues or establish that the tool was validated for your positions.
Can AI write employee handbooks or performance evaluations?
AI can help prepare a draft, but the employer should verify every factual statement, policy promise, and legal reference before using it. A handbook must reflect actual practices and applicable work locations. An evaluation should describe the employee’s real performance, not generic strengths or weaknesses generated by a tool.
Assign a reviewer who understands the subject. Remove invented incidents, inconsistent expectations, and language that conflicts with employment agreements or existing policies. Keep the approved version and a record of who authorized it.
Sources: NIST: AI Risk Management Framework.
Should an AI score determine a disciplinary or termination decision?
A score should not be accepted without understanding its inputs, limitations, and relevance to the employee’s work. Check whether protected leave, an accommodation, inaccurate data, or inconsistent management practices affected the result.
A responsible manager should review the underlying facts and document the actual reasons for the decision. Human approval alone does not cure a biased process. Where applicable law provides notice or reconsideration rights, those steps also need to be built into the workflow.
Sources: EEOC: Disability Discrimination; Colorado Attorney General: Automated Decision-Making Technology.
What should an employer check before adopting an AI employment tool?
Start with a written description of the business problem and the decisions the product will influence. Ask the vendor about testing, accessibility, data sources, security, retention, and changes to the model or scoring system.
Then assign internal ownership: who approves deployment, checks results, handles complaints, and can pause the tool. Test the actual configuration before use and review it after material changes. Preserve enough information to explain decisions later.
Sources: NIST: AI Risk Management Framework.
When should an employer seek legal advice about an AI tool?
Seek advice before deploying a tool that materially affects hiring, pay, promotion, discipline, or termination, and before agreeing to vendor terms that limit access to evidence. A complaint about bias, an inaccessible assessment, or unexplained differences in selection results warrants prompt review.
Bring the vendor agreement, product documentation, settings, work locations, and examples of the decisions involved. Those materials help counsel assess the actual process rather than the product’s marketing claims.
Updated October 21, 2026. General information for employers, not legal advice. Requirements depend on the facts, the employer, and the locations involved. Reading these FAQs does not create an attorney-client relationship.