The algorithm scored your resume and filtered it out before a human being ever saw it. The AI performance management platform ranked you in the bottom quartile. The automated scheduling tool repeatedly assigned you the worst shifts. A facial expression analysis system gave your interview a low score. A manager reviewed the recommendation, agreed with the output, and you were passed over, disciplined, or terminated. You never saw the data. You never knew the criteria. And your employer told you the decision was objective.

In California, that story has a legal answer. As of October 1, 2025, the California Civil Rights Department finalized sweeping new regulations under the Fair Employment and Housing Act (FEHA) that directly govern the use of artificial intelligence and automated decision systems in employment. These rules represent one of the most significant expansions of California employment discrimination law in years — and they apply to every employer in the state with five or more employees.

What the New AI Regulations Cover

On June 27, 2025, the California Civil Rights Council finalized amendments to FEHA regulations specifically targeting what the regulations call Automated Decision Systems, commonly referred to as ADS. An ADS is defined as any computational process — including artificial intelligence, machine learning, algorithms, statistics, or other data processing techniques — that makes a decision or facilitates human decision-making regarding an employment benefit.^[1]

The regulations, now codified at California Code of Regulations, title 2, sections 11008.1 through 11008.4, took effect on October 1, 2025, and apply to all California employers with five or more employees that use any such system in employment decision-making.^[2] The scope is deliberately broad. It includes tools used for resume screening, interview scheduling and scoring, skill or trait assessments, promotion recommendations, performance monitoring, scheduling algorithms, and any other computational process that influences what happens to you as a worker.

The core prohibition is simple: an employer may not use an ADS that discriminates against an applicant or employee, either intentionally or through disparate impact, based on any characteristic protected by FEHA — including race, national origin, sex, gender, gender identity, disability, religion, age, sexual orientation, and pregnancy.^[3]

Disparate Impact: Why Intent Does Not Matter

One of the most important legal principles in the new FEHA AI regulations is the disparate impact theory. Under this framework, you do not need to prove that your employer intended to discriminate. You do not need to show a manager made a biased decision. You need to show that the algorithm produced outcomes that disproportionately harmed members of a protected group — and that the employer cannot justify the system's criteria as a genuine business necessity.^[4]

This is particularly significant for AI discrimination cases because algorithmic bias is often invisible. A tool trained on historical data may encode the biases of past hiring decisions, producing results that systematically screen out women for engineering roles, filter out older workers for certain positions, or flag workers of certain national origins for more frequent monitoring — all without any human ever making a consciously discriminatory choice.^[5]

The regulations also address a subtle but powerful risk: the use of proxy characteristics. A proxy is a characteristic or category closely correlated with a protected class that, when used by an algorithm, produces discriminatory outcomes even though the protected characteristic itself is never explicitly considered. For example, an algorithm that screens out applicants from certain zip codes may effectively discriminate based on race if residential segregation patterns make zip code a proxy for race.^[6]

Employer Obligations: What the Law Requires

The FEHA AI regulations do not simply create liability — they create affirmative obligations. Employers must take proactive steps to ensure their AI tools are bias-free, and those steps carry their own legal significance. Conducting antibias testing or engaging in similar proactive efforts to avoid unlawful discrimination is specifically identified in the regulations as relevant evidence in support of defenses against discrimination claims. Conversely, the absence of antibias testing can itself be used as evidence against an employer.^[7]

The regulations also dramatically extend recordkeeping requirements. All ADS data — including input data, output scores, decision criteria, audit results, and related correspondence — must be preserved for at least four years. This doubles the previous two-year requirement and is designed to ensure that a discriminated-against worker who discovers the harm months or years later can still find evidence of what the algorithm did.^[8]

Critically, employer responsibility does not end at their own front door. The regulations expand the definition of "employer" under FEHA to include agents and third parties acting on behalf of an employer, including the vendors that develop or deploy automated decision systems. If your employer uses an AI hiring tool purchased from a software vendor, and that tool discriminates against you, your employer can still be held liable for the discrimination — even if they did not design the tool and did not know it was biased.^[9]

Workers must also receive notice before ADS tools are used in evaluating them, including an explanation of their rights and any opportunity to request human review or appeal the algorithmic outcome.^[10]

How to Prove an AI Discrimination Claim

Proving that an algorithm discriminated against you is a new legal frontier, but the tools are available. Under the four-year recordkeeping requirement, employers must retain the very data that can prove a discriminatory outcome. If you were passed over for a job, denied a promotion, or terminated after an ADS was used in the decision, your employer is required to have kept the algorithmic inputs and outputs that drove that decision.^[8]

Claims can be pursued under both the disparate treatment theory (the algorithm was programmed with discriminatory criteria or used demographic data directly) and the disparate impact theory (the algorithm disproportionately harmed a protected group without business justification). In either case, the employer's failure to conduct bias audits or document anti-bias efforts is powerful evidence against them.

California also enacted SB 53, the Transparency in Frontier Artificial Intelligence Act, effective January 1, 2026, which among other things adds whistleblower protections for employees of large AI companies who report concerns that a company's activities pose a specific and substantial danger to public health or safety. This reflects California's recognition that AI accountability must extend inside the companies building these tools as well as those using them.^[11]

The Bottom Line

Algorithmic discrimination is discrimination. California law does not allow employers to outsource their bias to a software vendor and call the result objective. The new FEHA AI regulations make clear that if an automated system screens you out, terminates you, or denies you opportunities based on criteria that disproportionately harm your protected group — even without any human making a conscious biased choice — your employer may be fully liable under FEHA.

At Smith Reback Law, we represent California workers navigating this new frontier of employment discrimination law. Whether you were filtered out by a resume screening algorithm, scored poorly by an AI interview tool, flagged for termination by a performance management platform, or denied advancement by predictive analytics, we know how to build these cases, demand the algorithmic records, and hold employers accountable. Your consultation is completely confidential and costs nothing.