AI hiring in California: Can employers defend the outcomes?

Imagine defending a discrimination lawsuit in which every hiring decision from the past three years has been scored, ranked and stored by an artificial intelligence platform. Defending such a lawsuit is no longer hypothetical, as employers use AI for all stages of the employment lifecycle. AI can screen resumes, rank applicants, analyze video interviews, identify candidates for promotion, evaluate performance and support workforce planning. 

Two years ago, the EEOC filed a lawsuit against iTutorGroup, three integrated companies that provided English-language tutoring services to students in China. The EEOC alleged that iTutorGroup violated the ADEA when it programmed tutor application software to automatically reject all female applicants aged 55 or older and male applicants aged 60 or older. As a result, iTutorGroup's $365,000 settlement was distributed to about 200 rejected (but qualified) applicants.  

But what if the company did not intentionally program the software to reject qualified applicants based on discriminatory criteria like age? Or what if the company licensed AI from a third-party and was not involved--or even aware of--its programming decisions? Employers are likely responsible for discriminatory outcomes even when screening decisions are delegated to automated systems or third-party vendors. 

Of course, the legal framework for discriminatory outcomes is not new. Decades ago, the United States Supreme Court established the concept of disparate impact in Griggs v. Duke Power Co., holding that an employer could be liable for facially neutral hiring requirements that nevertheless disproportionately excluded Black applicants. AI-driven hiring tools pose a modern version of the same risk: An algorithm may apply the same criteria to every applicant, yet still produce unlawful outcomes if those criteria disproportionately disadvantage protected groups and cannot be justified by business necessity. 

California is once again at the forefront of regulating technologies used to make employment decisions. On June 30, 2025, the California Civil Rights Council announced that new rules to protect against potential employment discrimination as a result of AI, algorithms and other automated decision systems (ADS) would become effective Oct. 1, 2025. The new rules make it clear that employers cannot assume technology is neutral simply because it is automated. The regulations governing ADS focus on whether those systems produce discriminatory outcomes, not on whether anyone intended discrimination. If an AI tool disproportionately disadvantages applicants or employees based on race, gender, age, disability or another protected characteristic, employers may face liability under California's Fair Employment and Housing Act (FEHA). 

The new rules also put the onus on employers in defending a claim to garner evidence demonstrating that anti-bias testing or similar proactive efforts have been employed to prevent unlawful discrimination, including "the quality, efficacy, recency and scope of such effort, the results of such testing or other effort, and the response to such results."   

Employers subject to California regulatory scheme will need to determine how the AI tools that they are using in employment decisions are operating. Relying on a vendor's assurances that a platform is compliant or "bias-free" will offer little comfort if a regulator or plaintiff later identifies a disparate impact on protected groups. 

Thus, a meaningful bias audit is an important risk-management tool. For example, an employer using AI to screen resumes might compare interview advancement rates across demographic groups each quarter. Audits may need to occur as vendors are changed or internal systems are materially modified. The process of documenting methodologies, findings and remediations is critically important from a litigation perspective. A defensible audit goes beyond a vendor certification: It examines disparities, tests whether the criteria are job-related, considers less discriminatory alternatives and records the employer's response.

As AI becomes more prolific, so too will be the scope of what an employer should expect to produce in discovery. Employers should expect requests for audit reports, validation studies, vendor communications, internal analyses, meeting minutes and records of bias concerns. The question will not be simply whether disparities existed, but whether the employer knew (or should have known) about them and what it did in response.

Recent litigation against Workday illustrates how the landscape will change when it comes to what is discoverable in AI-related employment cases. In Mobley v. Workday, Derek Mobley filed a putative class action alleging that Workday's AI-based applicant recommendation system, which it provided to its customers, discriminated against job applicants on the basis of race, age and disability. During discovery, plaintiffs sought production of Workday's bias-testing and applicant data. Workday successfully argued that its bias-testing was privileged because its attorneys curated the data, and the data was not used in making business decisions or submitted to regulators. Plaintiffs also were unsuccessful in convincing the court that Workday should be compelled to produce applicant data belonging to its customers. That said, the court acknowledged some third-party customers had taken the position that Workday was the better source, leaving this issue open for another day. Finally, the court ordered the production of Workday's EEO-1 and OFCCP documents because those documents were relevant to Workday's knowledge of potential demographic disparities when using AI tools.  

Artificial intelligence can be a powerful business tool. It can improve efficiency, streamline hiring and support workforce decision-making. However, it is not a substitute for human oversight, compliance or sound judgment. 

The employers best positioned for the future will treat AI not as a technology initiative but as a governance issue requiring ongoing monitoring, documentation and accountability. In California's evolving regulatory environment, the question is no longer whether your organization uses AI in employment decisions. It is whether you can defend the outcomes.