California’s Automated-Decision Rules Under FEHA
Since October 1, 2025, California’s Fair Employment and Housing Act regulations explicitly cover automated-decision systems. They create no new audit mandate, but they change what counts as evidence and who can be liable.
What an automated-decision system is
Any computational process, including AI, machine learning or other algorithms, that makes or helps make decisions about applicants or employees. Resume screening, interview scoring, skill or trait assessments and promotion recommendations are in scope. Ordinary IT tools such as email, word processors and spreadsheets are not.
Three changes that matter
- Discrimination through an ADS is unlawful. It is an unlawful employment practice to use an automated-decision system or selection criteria that discriminate on a protected basis.
- Vendors can be liable as agents. The rules define an employer’s agent to include third parties acting on its behalf, such as staffing agencies and vendors. Employers can be responsible for disparate impact caused by a tool a vendor runs for them.
- Testing is evidence. Anti-bias testing and similar proactive efforts are relevant to a claim or defense. Courts and the agency look at the quality, efficacy, recency and scope of the effort, the results, and how the employer responded.
Record keeping
Employers must keep automated-decision data, including inputs, outputs, criteria and testing results, for at least four years, and longer if a complaint is filed.
How it interacts with the CCPA rules
Separately, California’s privacy regulator has adopted automated-decision rules under the CCPA. For businesses covered by the CCPA, they require a pre-use notice, an opt-out (with a human-appeal exception), access rights and a written risk assessment before using automated decision-making for significant employment decisions, with compliance due by January 1, 2027.
What to do
- Inventory every tool that scores, ranks or recommends Californian applicants or employees.
- Test each for adverse impact before use and on a schedule, and document what you did about the results. See the four-fifths rule, worked through.
- Review vendor contracts for testing cooperation and data access.
- Set four-year retention for ADS inputs, outputs and test results.
- Keep accommodation routes open for candidates who cannot use an AI assessment.
Map your hiring AI against the laws that apply to you
Whether you build HR AI or deploy it, Fairly AI tests it, monitors the rules where you operate and keeps the evidence.
Schedule a discovery call