Employment Law

Using AI in Hiring: California's Automated Decision System Rules

By Ingrid Rainey · August 4, 2026 · 7 min read

Editorial photo illustrating Using AI in Hiring: California's Automated Decision System Rules

Resume screeners, video interview scoring, personality assessments, and predictive attrition models are now squarely within California's employment discrimination framework. The Civil Rights Department's regulations make clear that using an automated decision system that produces a disparate impact on a protected group is unlawful discrimination by the employer, whether or not the employer understood how the tool worked. The vendor's marketing claims are not a defense.

Key takeaways

  • The definition is broad: computational processes that make or substantially facilitate an employment decision.
  • Proxy variables — zip code, school, employment gaps, or graduation year — correlate with race, national origin, and age.
  • Inventory every tool touching recruiting, screening, promotion, scheduling, and discipline, including features embedded in your applicant tracking system.
  • California's rules reach agents acting on an employer's behalf, so a staffing agency or screening vendor can be liable alongside the employer.

What counts as an automated decision system

The definition is broad: computational processes that make or substantially facilitate an employment decision. That includes resume-ranking software, keyword filters, game-based assessments, facial or voice analysis in interviews, targeted job-advertising algorithms, and scheduling or productivity tools that feed discipline decisions.

Substantially facilitate matters. An employer that reviews every candidate but sees them in an algorithmically ranked order is still using the system to shape the outcome.

Where these tools go wrong

  • Proxy variables — zip code, school, employment gaps, or graduation year — correlate with race, national origin, and age.
  • Video and voice analysis can penalize accents, speech differences, and visible disabilities, implicating both national origin and disability discrimination.
  • Personality and health-adjacent assessments can constitute a prohibited medical or psychological inquiry before a conditional offer.
  • Systems trained on a company's historical hires reproduce the historical composition of that workforce.
  • No accommodation pathway: candidates who need an alternative format must be able to request one, and the process must actually deliver it.

Have a question about your situation?

Rainey Law P.C. advises San Diego businesses and employees on matters like this one.

Request a consultation

The employer's compliance checklist

  • Inventory every tool touching recruiting, screening, promotion, scheduling, and discipline, including features embedded in your applicant tracking system.
  • Demand bias-testing documentation from each vendor, and treat the absence of testing as evidence against the tool, not as neutral.
  • Run your own adverse impact analysis on outcomes by protected group, under privilege, and repeat it periodically.
  • Confirm each screening criterion is job-related and consistent with business necessity, and that no less discriminatory alternative would serve the same need.
  • Publish an accommodation notice in the application flow and staff the request path.
  • Negotiate contract terms: audit rights, data access, indemnity, and a representation that the tool has been tested for adverse impact.
  • Retain applications, personnel records, and automated-decision data — including the criteria and outputs — for at least four years.

Vendor contracts and shared liability

California's rules reach agents acting on an employer's behalf, so a staffing agency or screening vendor can be liable alongside the employer. That does not shift the employer's own exposure. Practically, the employer is the party a claimant sues, and the vendor's indemnity is only worth its cap and its balance sheet.

Before renewing a hiring-tech contract, align the indemnity with the realistic scale of a class or agency action, and preserve the right to obtain the data you would need to defend a claim.

Rainey Law P.C. advises employers on AI hiring governance, vendor terms, and privileged adverse impact analysis.

Frequently asked questions

Is it illegal to use AI in hiring in California?

No. Using an automated decision system is lawful; using one that produces a disparate impact on a protected group without job-related justification is not. The obligation is to test, document, and correct.

Our vendor says the tool is bias-free. Is that enough?

No. Liability follows the employment decision, not the software license. Ask for the actual testing methodology and results, and validate against your own applicant pool.

What records do we need to keep?

Applicant and personnel records, plus the automated-decision data, criteria, and outputs, generally for four years. Missing records make an adverse impact claim much harder to defend.

This article is general information, not legal advice, and does not create an attorney-client relationship. Outcomes depend on the facts of each matter.

More insights

We fight for you.

Schedule a confidential consultation to discuss your business or employment matter.

Call NowBook Consult