Skip to content
California’s FEHA automated-decision rules: vendor and employer exposure
California’s civil-rights rules on automated-decision systems have applied to hiring since October 2025. No audit is mandated, but testing is a legal defence, and liability reaches the vendors that run screening, not just the employer.
Last reviewed: August 2026
Adverse impact testing extended to California’s FEHA protected classes, with the records and documentation that support an employer’s defence. Covers one AI hiring tool.
Request this testFixed fee, one tool. Fees exclusive of VAT where applicable. Covering more than one US rule? Our combined US report does all four for $3,250.
What the FEHA rules require
Since 1 October 2025 the California Civil Rights Council’s regulations under the Fair Employment and Housing Act apply to automated-decision systems used in employment, for employers with five or more staff. Using a system that discriminates, by disparate impact or intent, on a protected basis is unlawful.
Liability is not limited to the employer. The rules extend to an employer’s agents, which expressly includes third parties that carry out recruitment, screening or hiring functions. That puts HR-technology vendors, not just their customers, in scope.
No bias audit is mandated. But anti-bias testing, and its quality, recency and scope, is expressly relevant evidence, and the absence of it can be used against you. Records must be kept for four years. Separately, the California Privacy Protection Agency’s automated decision-making technology rules add notice and opt-out duties, with compliance required by 1 January 2027.
Testing, evidence and vendor cover
Extended-class testing
Adverse-impact testing across the FEHA protected classes, broader than the categories a New York City audit covers.
Vendor and agent testing
For HR-tech providers: independent testing of your tool, because the agent rule can put you directly in scope, not just your customers.
Evidence and records
Documented testing and results to serve as a defence, kept to the four-year retention the rules expect.
FEHA protected classes a bias analysis should consider include race, colour, ancestry, national origin, religion, age (40+), disability, medical condition, genetic information, sex, gender, gender identity and expression, sexual orientation, marital status, pregnancy and related conditions, military and veteran status, and reproductive health decision-making. California adds bases from time to time, so confirm the current list before relying on it.
Note: SB 947, the “No Robo Bosses” bill, has passed the California Senate but is not yet law as of August 2026. We are watching it.
This page is general information, not legal advice. AI employment law is moving quickly; confirm your position with qualified counsel before relying on it.
Build or use hiring AI in California?
The agent rule can put the vendor on the hook, not just the employer. Independent testing is how you show the tool does not discriminate.
Compare US jurisdictions