California FEHA Automated-Decision-System Hiring Regulations 2026
NYC's Local Law 144 and Illinois's HB 3773 get most of the attention, but California quietly did something bigger: it wrote AI hiring tools directly into its existing anti-discrimination statute. FEHA's Civil Rights Council regulations don't create a new AI law — they make clear the old discrimination law already covers your ATS, your screening algorithm, and your interview-scoring model.
Why This Is a Different Kind of AI Law
Most of the AI-hiring-law conversation centers on new statutes: NYC's bias-audit mandate, Illinois's AI notice requirement, Colorado's AI Act. California took a different route. Rather than pass a new AI-specific law, its Civil Rights Council amended the regulations implementing FEHA — the state's existing employment discrimination statute — to explicitly address automated-decision systems.
The practical effect: there's no separate "AI hiring law" to check off a compliance list and move on. An AI hiring tool that produces a disparate impact on a protected class is now unambiguously a FEHA violation, subject to the same enforcement, the same uncapped damages, and the same individual right to sue that any other discriminatory hiring practice carries.
What Counts as an Automated-Decision System
The definition is intentionally broad — it reaches AI, machine learning, algorithms, and statistical models used to make or assist an employment decision. This isn't limited to fancy generative-AI screening tools:
Resume screening and ranking software
Keyword-matching or scoring tools that filter or rank applicants before a human sees them are squarely covered, even if the underlying tech is simple rules rather than machine learning.
Video interview and personality/skills assessments
Tools that score facial expressions, speech patterns, or assessment responses to predict job fit are exactly the category that's drawn prior EEOC and state scrutiny for disparate impact.
Chatbot-based candidate screening
Conversational pre-screen bots that decide who advances to a human interview are an ADS, whether the deciding logic is an LLM or a scripted decision tree.
Workforce-management systems used in discipline or scheduling
The regulations aren't limited to hiring — systems used in promotion, discipline, termination, or scheduling decisions with adverse impact potential are also covered.
Recordkeeping: The Duty Most Employers Miss
FEHA's employment recordkeeping rules extend to ADS-related records: the tool used, the criteria or model logic where available, and the data behind the employment decision. The practical floor is four years from the date of the record or the employment action, whichever is later relevant. Employers that use a vendor's black-box scoring tool and don't retain the inputs, scores, or decision rationale have no way to defend a disparate-impact claim later — the absence of records itself becomes evidence against them.
Vendor Liability Doesn't Replace Employer Liability
A common but incorrect assumption: "our ATS vendor is responsible for making sure the algorithm is fair." FEHA liability attaches to the employer that used the tool to make an employment decision, full stop — a vendor's own potential liability as an "agent" of the employer is additive, not a substitute. Contract language disclaiming the vendor's responsibility for bias does not transfer legal exposure away from the employer using the tool on California applicants or employees.
Compliance Checklist for California Employers
- Inventory every ADS touching hiring, promotion, discipline, or termination decisions for California-based roles.
- Request or run bias/adverse-impact testing from vendors before deployment, and periodically thereafter.
- Retain inputs, scores, and decision logic for at least four years, not just the final hire/no-hire outcome.
- Update vendor contracts to require bias-testing disclosure and audit cooperation — but don't rely on contract terms to shift legal liability.
- Train recruiters and HR on what counts as an ADS in your stack; shadow-IT tools adopted by individual recruiters are a common blind spot.
- Document a human-review step for adverse decisions driven substantially by an ADS score.
See where your AI product is exposed
RatedWithAI helps SaaS teams surface compliance and trust gaps across their web properties. Start with a free scan to understand how your product presents to users and regulators alike.
Scan Your Product for Free →Frequently Asked Questions
How is this different from NYC's Local Law 144?
LL 144 requires an independent annual bias audit and public disclosure before using an automated employment decision tool. California's FEHA regulations don't mandate a specific audit format — instead, they fold ADS use into existing disparate-impact liability, with bias testing serving as evidence of good faith rather than a standalone filing requirement.
Does this apply to small businesses?
FEHA generally applies to employers with 5 or more employees, which is a much lower threshold than many federal discrimination statutes. Most California employers using any AI hiring tool are covered.
What if our AI vendor won't share bias-testing data?
That's a red flag, not a shield. If a vendor won't provide adverse-impact testing results or explain the model's decision logic, the employer bears the litigation risk of deploying an unauditable tool on California applicants — treat vendor transparency as a procurement requirement, not a nice-to-have.
Are damages capped like some other state AI laws?
No. FEHA claims carry the same uncapped compensatory and punitive damages, attorney's fees, and individual right to sue as any other California employment discrimination claim — there's no AI-specific damages cap.