California AI Hiring Rules: Employer Guide

California’s new rules on automated employment decisions require employers to audit their AI tools, notify workers, and give them the right to appeal. This guide explains what the law covers, how to comply, and what happens if you don’t.

Background: Why California Regulates AI in Hiring

California has long led U.S. privacy and employment law, from the California Consumer Privacy Act (CCPA) to strict anti-discrimination statutes. The state’s regulations on automated decision-making in employment stem from a growing concern: algorithms used in hiring, promotion, and performance reviews can unintentionally discriminate, even when designed to be neutral. In 2020, California passed a law requiring employers to report pay data by race and gender, and in 2022, it amended the CCPA to give consumers—including job applicants and employees—the right to opt out of automated decision-making in certain contexts. However, the most direct rule for employers is the California Privacy Rights Act (CPRA) regulations, which took effect in 2023 and specifically address “automated decision-making technology� (ADMT) in employment.

The state’s Fair Employment and Housing Act (FEHA) also applies: if an AI tool has a disparate impact on a protected class, the employer can be liable even without intent to discriminate. In 2024, California’s Civil Rights Department issued guidance clarifying that employers must conduct “bias audits� for AI tools that screen or rank applicants. While a federal law on AI hiring has not passed, California’s rules are among the strictest in the nation, and they apply to any employer with workers in California, regardless of where the company is headquartered.

What the Law Requires: Key Obligations

Under the CPRA regulations, employers using automated decision-making technology for employment purposes must provide:

• Pre-use notice: Before using an AI tool to make a decision that significantly affects employment (e.g., hiring, firing, promotion, pay), the employer must inform the applicant or employee that the tool is being used, what data is collected, and how the decision will be made.
• Right to opt out: For certain types of decisions, the individual must have the right to request that the employer not use the automated system and instead use a manual alternative. This applies to decisions like “significantâ€� ones—hiring, promotion, termination, and compensation.
• Access and correction: The individual has the right to access the data used in the decision and to correct inaccurate data.
• Right to appeal: If the automated decision results in an adverse action (e.g., rejection, demotion), the individual must have a process to appeal and have a human review the decision.
• Bias audit: Employers must conduct a bias audit of the AI tool before using it, and again at least once every two years. The audit must assess whether the tool has a disparate impact on protected groups (race, gender, age, disability, etc.).

These requirements apply to tools that use “machine learning� or “artificial intelligence� to make or inform decisions. Simple spreadsheets or manual scoring systems are not covered, but any automated system that ranks, scores, or filters candidates is likely included.

How to Comply: Steps for Employers

To comply with California’s rules, employers should take the following steps:

1. Inventory your tools: Identify every software or system that uses automated decision-making in employment—resume screeners, video interview analysis, personality tests, performance monitoring, and even scheduling algorithms.
2. Conduct a bias audit: Use a qualified auditor or internal team to assess the tool’s impact on protected groups. The audit should include statistical analysis of outcomes (e.g., selection rates by race, gender, age) and a review of the tool’s design and training data.
3. Update privacy notices: Revise your applicant privacy policy to clearly describe the ADMT used, the categories of data collected, and the purpose.
4. Implement an opt-out process: For decisions where opt-out is required, provide a simple way for individuals to request manual review. This could be an email address or a form in your applicant tracking system.
5. Create an appeal procedure: When a candidate is rejected or an employee is demoted or terminated based on an automated decision, notify them in writing and explain how to appeal. The appeal must be reviewed by a human with authority to change the decision.
6. Document everything: Keep records of your audits, notices, and appeals for at least two years, as required by the CPRA.

Common Problems and Misconceptions

One common misconception is that the law only applies to large tech companies. In reality, it applies to any employer with at least one California employee or applicant, regardless of company size. Another misconception is that using a third-party vendor absolves the employer of responsibility. Under the CPRA, the employer is the “business� and is responsible for the vendor’s compliance unless the vendor is a service provider with a contract that limits its use of the data. Employers often struggle with the bias audit requirement because there is no standard methodology. The California Privacy Protection Agency (CPPA) has proposed rules that would require audits to be conducted by a “qualified individual� and to include specific metrics, but as of 2025, the final rules are still pending. In the meantime, employers should follow the CPPA’s draft guidance and the Fair Employment and Housing Council’s regulations, which suggest using the “four-fifths rule� (a selection rate for a protected group that is less than 80% of the highest group’s rate suggests adverse impact) but also allow for more sophisticated statistical tests.

Another practical problem: many AI hiring tools are “black boxes�—the vendor doesn’t disclose how the algorithm works. Employers must demand transparency from vendors, including details about training data and feature weights, to conduct a meaningful audit. If the vendor refuses, the employer may need to choose a different tool.

Comparison with Other States and Federal Law

California is not the only state addressing AI in hiring. Illinois has the Artificial Intelligence Video Interview Act, which requires notice and consent for video interviews analyzed by AI. New York City has a law requiring annual bias audits for automated employment decision tools, with penalties for non-compliance. Maryland and New Jersey have similar laws. Federal law, including Title VII of the Civil Rights Act and the Equal Employment Opportunity Commission’s (EEOC) technical guidance on AI, prohibits discrimination but does not yet require audits or notice. California’s rules are broader because they apply to all employment decisions, not just hiring, and they include the right to opt out and appeal, which other states lack. For employers operating in multiple states, the safest approach is to follow the most stringent requirements—typically California’s—for all applicants and employees, or at least for those in California.

Costs and Penalties for Non-Compliance

Compliance is not free. A bias audit can cost anywhere from a few thousand dollars for a small tool to tens of thousands for a complex system, depending on the auditor and the data involved. Updating privacy notices and training HR staff also takes time and money. However, the cost of non-compliance is higher. The CPRA allows for administrative fines of up to $2,500 per unintentional violation and $7,500 per intentional violation. Civil penalties under FEHA can reach $150,000 per violation for discriminatory practices. Additionally, class-action lawsuits are a real risk—if a rejected applicant can show that an AI tool disproportionately screened out a protected group, the employer could face significant damages. Beyond legal costs, there is reputational harm: companies like Amazon and Uber have faced public backlash over biased hiring algorithms.

Frequently Asked Questions

Does California’s rule apply to small businesses?

Yes, there is no minimum employee threshold. Even a small business with one California employee must comply if it uses automated decision-making technology.

What counts as an “automated employment decision�?

It includes any decision that significantly affects an individual’s employment, such as hiring, promotion, termination, compensation, and work assignment. It does not include routine decisions like scheduling shifts unless the schedule is based on algorithmic predictions of performance.

Can we use an AI tool that doesn’t allow for an opt-out?

No. If the tool is used for a significant decision, you must provide an opt-out. If the tool cannot accommodate a manual review, you cannot use it for that purpose in California.

How do we conduct a bias audit if we don’t have a data scientist?

You can hire an external consultant or use a software tool designed for bias detection. The law requires the audit to be “reasonable� and based on available data. If you lack the expertise, it’s best to hire a professional.

What if our vendor won’t share details about the algorithm?

You are still responsible for compliance. If the vendor won’t provide sufficient information for an audit, you should stop using the tool in California or find a vendor that is transparent.

Does this apply to current employees or only job applicants?

Both. The rules cover all employment decisions, including promotions, performance evaluations, and terminations for existing employees.

Practical Tips for Getting Started

Start by reviewing your current hiring and HR software. If you use any tool that automatically screens resumes, ranks candidates, or analyzes video interviews, you likely need to take action. Prioritize the tools that have the most significant impact on decisions. Document your compliance efforts, as this can mitigate penalties if a violation is found. Finally, stay informed: the CPPA is still finalizing regulations, and California’s legislature may pass new laws, so check for updates annually.

Related

• Used Sports Sedan: BMW Power vs Lexus Luxury
• Kobalt Power Tools at Lowe’s: Worth Buying?
• Cheap Power Tools Worth Buying? Pro Picks