Effective October 1, 2025, updated regulations from the California Civil Rights Council will formally restrict the use of artificial intelligence (AI) tools in employment decision-making by California employers. In the employment context, these tools can be applied in a litany of ways to manage the workforce, including to screen resumes, make predictions about an applicant or employee, measure an applicant or employee’s skills or abilities, direct job advertisements and recruiting materials to targeted groups, and screen, evaluate, and/or recommend applications or employees.
The New Rules Apply Existing Anti-Discrimination Law to AI Tools
Rather than devise an entirely new set of regulations concerning AI, the Civil Rights Council decided to amend the existing regulatory framework of the Fair Employment and Housing Act (FEHA) to any “automated decision system.” As such, the AI regulations functionally treat AI as another source of potential impermissible discrimination based on protected characteristics. Ordinarily, such a tactic would be less-than-noteworthy since caselaw interpreting the FEHA is well-established, but in this instance, the Civil Rights Council’s updated regulations present significant changes, including:
- For the first time, defining an employer’s “agent” as anyone acting on behalf of an employer – directly or indirectly – “to exercise a function traditionally exercised by the employer or any other FEHA-regulated activity.” The regulations specifically identify applicant recruitment, screening, hiring, promotion, or decisions regarding pay, benefits, or leave as potential examples of such “function,” and specify that an automated decision system can qualify as an agent of the employer.
- Broadly defining an automated decision system as “[a] computational process that makes a decision or facilitates human decision making regarding an employment benefit.” The updated regulations consider AI, along with “machine-learning, algorithms, statistics, and/or other data processing techniques” to be automated decision systems.
- Defining AI as “a machine-based system that infers, from the input it receives, how to generate outputs,” which is not a commonly-accepted description of AI, but is instead much closer to the definition of a computer. As currently drafted, this extremely broad definition of AI has the potential to include any computer program within its scope, even if such a program does not perform complex tasks normally done by human-reasoning.
- Establishing a new and distinct affirmative defense for employers that use automated decision systems in FEHA lawsuits: evidence of anti-bias testing or similar efforts to avoid unlawful discrimination.
- Requiring employers to keep records of its automated decision systems decision-making data for at least four years.
The updated rules are not aimed only at explicit bias: they also apply to algorithms that have an adverse impact on a protected group, often referred to as having a “disparate impact.” Thus, even if an employer is not intentionally using AI to discriminate among applicants or employees, the new rules prohibit the use of AI where a disparate impact results.
Employers Have Multiple Options to Mitigate the Risk of Running Afoul of the Updated Civil Rights Council Regulations
The October 1, 2025 effective date of the updated regulations provide California employers with a few months to confirm that their use of AI – or any computer systems – in the context of workforce management conforms to the law. In light of the updated regulations’ formal recognition of the anti-bias training affirmative defense, employers should consider focusing their efforts on reducing the risk that their use of AI results in disparate impact based upon protected characteristics. As recognized in the updated rules, there are various means by which risk reduction can be effectuated, such as:
- Conducting bias audits or impact assessments before implementing AI technology;
- Regularly reviewing the impact of AI tools on protected groups after implementation;
- Evaluating the quality of audits, assessments, and/or tests used to detect potential discriminatory outcomes in the use of AI tools;
- Documenting the results of any anti-bias testing, including the employer’s response to the results; and
- Instituting human review of any employment decisions facilitated by AI.
The New Rules’ Impact on Workforce Management Vendors
Because the updated regulations broadly define an agent of an employer, such that the owner or licensor of an automated decision system may qualify as an agent of an employer under the FEHA, vendor management will play a key role in mitigating the risk associated with the use of AI. California employers should carefully review their indemnification and defense agreements with vendors, while also mandating that their vendors certify the efficacy and results of their anti-bias testing conducted on AI platforms. The use of AI in workforce management presents an exciting opportunity for California employers, but such use is fraught with risk if not accompanied with a diligent eye towards legal compliance and appropriate risk balancing.