Artificial intelligence (AI) is no longer just a future-facing issue for California employers. Many businesses already use some form of automated tool to sort résumés, rank applicants, target recruiting, score assessments, flag productivity concerns, or assist with promotion and compensation decisions. What is different now is that California’s employment regulations addressing automated-decision systems are in effect, which makes this a present compliance issue rather than a speculative one. These regulations went into effect on October 1, 2025 .

The key point for employers is simple: using AI does not insulate an employment decision from scrutiny under California law. The regulations are designed to clarify how existing anti-discrimination rules apply when an employer uses artificial intelligence, algorithms, or other automated-decision systems in employment decisions.

The regulations also define the issue broadly. They are not limited to a futuristic chatbot or a fully autonomous hiring machine. The final text addresses “automated-decision systems” used to make or facilitate employment decisions, and the regulatory materials have described examples such as résumé-screening tools, targeted recruiting systems, computer-based assessments, and technologies that analyze facial expression, voice, or word choice in online interviews.

That breadth matters because many employers may not even think of their existing human resources (HR) software as “AI.” A vendor may market a tool as analytics, workflow optimization, or talent intelligence, but the legal question is not the product label. The real question is whether the system helps screen, rank, recommend, categorize, or otherwise influence employment decisions in a way that could disadvantage applicants or employees based on protected characteristics. The California rules are aimed at that functional reality.

Another important point is that vendor involvement does not eliminate employer risk. California’s regulatory framework treats certain third parties acting on an employer’s behalf as agents, which means employers should not assume that outsourcing the technology also outsources liability. If an employer is relying on a third-party product in recruiting, hiring, promotion, discipline, or related decision-making, vendor diligence becomes part of employment-law compliance.

So, what should California employers be asking now?

1 – What Automated Tools Are We Already Using Affecting Employment Decisions?

The first step is inventory. Employers often focus on applicant screening, but the risk points can be much broader. The federal U.S. Equal Employment Opportunity Council (EEOC) has identified hiring, employee monitoring, pay-setting, promotion, and termination as areas where AI and automated systems may implicate discrimination law. California’s regulations likewise address automated systems used in employment decision-making more generally.

In practice, that means employers should look beyond recruiting software. Performance-management tools, productivity trackers, scheduling platforms, assessment vendors, and internal ranking systems may all deserve review if they meaningfully influence decisions about applicants or employees. That review is especially important where the system is treated as objective or neutral simply because it is software-driven.

2 – Are We Using Criteria That Attack Protected Traits?

A recurring legal risk in this area is proxy discrimination. A system may rely on criteria that appear facially neutral but still correlate with race, disability, sex, age, or another protected characteristic. California’s regulatory text addresses “proxy” concepts in this context, and that should prompt employers to look closely at the inputs and outputs of any automated screening or ranking tool.

That means asking hard questions about what the system actually measures. Is it drawing inferences from gaps in work history, speech patterns, facial movements, location data, educational pedigree, or other variables that may create a skewed outcome? If the answer is unclear, that itself is a warning sign. Employers should be wary of relying on a candidate “score” they cannot explain – let alone, identify how it is being calculated.

3 – Are We Creating Disability Discrimination?

This is one of the most significant areas of AI-related employment risk. Both California’s Civil Rights Division (CRD) and the EEOC have emphasized that technology can create discrimination issues for applicants and employees with disabilities. The EEOC has given examples involving video-interviewing software that scores applicants based on speech patterns and facial-analysis systems that may not accurately assess individuals with certain disabilities.

For California employers, that means AI-assisted assessments, video interview tools, testing platforms, and productivity-monitoring systems should be reviewed not only for general bias concerns, but also for disability accommodation issues and disability-related inquiry concerns. An employer may still face liability if a system screens out qualified individuals with disabilities or elicits information in a way that violates existing law.

4 – Are We Using Meaningful Human Review, or is the Software Making the Call?

Many employers assume that a human “in the loop” solves the problem. Often it does not. If the human reviewer is merely rubber-stamping the algorithmic recommendation, the software may still be doing the real work. The better question is whether the reviewer actually understands the basis for the recommendation, can identify when the output looks unreliable, and has authority to depart from it. This is a practical inference from the regulations’ focus on systems that make or facilitate employment decisions.

That makes training important. Supervisors, recruiters, and HR personnel should understand that an automated result is not inherently defensible because it came from a vendor platform. A recommendation generated by software is still something the employer may later have to justify.

5 – Are We Preserving the Data and Can We Explain What Happened?

This may be the most overlooked compliance issue. Employers and other covered entities must maintain employment records, including automated-decision-system data, for at least four years. For employers using AI-related tools, recordkeeping may become critical in any future administrative charge or litigation.

If a challenged employment decision turns in part on an automated score, ranking, recommendation, or screening result, the employer may need to explain what tool was used, what information went into it, what output it generated, who reviewed it, and what role it played in the final decision. If those facts are not preserved, defending the decision becomes harder.

Practical Steps for California Employers:

California employers do not need to abandon technology. But they should treat AI governance as an employment-law issue, not merely an IT or procurement issue. A sensible starting point is to inventory existing tools, identify where those tools influence employment decisions, review vendor contracts and diligence materials, confirm whether there is a process for accommodations and overrides, and make sure the business can preserve the relevant records. Those steps align with the direction of California’s regulations and with the EEOC’s repeated warnings that existing discrimination laws still apply when AI is involved.

The bottom line is that California employers should stop asking whether AI is “allowed” and start asking whether their actual use of automated tools can be explained, monitored, and defended under existing employment laws. As of April 2026, that is not an emerging issue. It is already here.

For any questions or assistance with coming into compliance with the new AI statutes and regulations, please contact us at Carle Mackie Power & Ross LLP. Arif Virji, Justin D. Hein, Samantha Pungprakearti, Sarah Hirschfield-Sussmann.