California's New AI Workplace Laws: What Employers Must Do Before 2027

Last year, Governor Newsom vetoed California's first “No Robo Bosses” bill, SB 7, in October 2025. This year he signed its successor. On September 30, 2026, he approved a set of workplace AI bills that will change how California employers use software to discipline, fire, lay off, and monitor people. The Governor's announcement lists them alongside roughly a dozen other AI measures.

If you have even one employee in California, at least one of these laws applies to you. Most small employers don't think of their software as “AI,” and that is the first thing to fix. The definitions are broad enough that they can pick up scheduling apps, productivity dashboards, and the scoring features built into many HR platforms.

This piece is organized by date, because that is how you'll have to work through it.

The four laws in one paragraph each

SB 947, the No Robo Bosses Act (Statutes of 2026, ch. 859). You may not rely solely on an automated decision system to discipline or fire an employee. If you rely primarily on one, a human has to corroborate the output, and you owe the employee a written notice. Operative July 1, 2027.

SB 951, AI-related layoff notices (Statutes of 2026, ch. 860). This amends California's WARN Act. If a mass layoff, relocation, or termination is caused “in whole or in substantial part” by AI, your WARN notice has to say so and describe the technology. Effective January 1, 2027.

AB 1883, AI workplace surveillance (Statutes of 2026, ch. 853). You may not use an AI-based workplace surveillance tool to infer or predict employees' emotional states or to collect their neural data, with an exception for use “to ensure safety.”

AB 1331, workplace surveillance (Statutes of 2026, ch. 851). This one is narrower and not AI-specific. It bars using workplace surveillance tools to monitor employees in bathrooms, with limited exceptions.

AB 1883 and AB 1331 don't state their own start dates. Under California's default rule for non-urgency statutes, they take effect January 1, 2027.

Who is covered (the thresholds question)

The answer differs by law, and it is the question small employers ask first.

• SB 947 has no headcount threshold. Its definition of “employer” covers anyone who employs someone, directly or through an agent. It expressly includes state and local government and labor contractors. A five-person shop is covered the same way a five-thousand-person company is. It protects employees. It does not reach applicants, and the hiring stage is governed by the other rules discussed below.
• SB 951 rides on Cal-WARN's existing threshold. Cal-WARN applies to a “covered establishment” that employs, or has employed in the preceding 12 months, 75 or more people. The EDD's WARN page explains how the count works. The triggering events are unchanged: a mass layoff of 50 or more within 30 days, a closure, or a relocation of at least 100 miles. Under 75, SB 951 doesn't touch you.
• AB 1883 and AB 1331 have no headcount threshold in their text. Both define “employer” broadly, including public employers and labor contractors.

January 1, 2027: layoff notices and surveillance limits

SB 951: the “technology displacement” notice

Cal-WARN already requires 60 days' notice to affected employees, the EDD, and local officials before a covered layoff, relocation, or closure. SB 951 adds content to that notice when the event is caused in whole or in substantial part by an AI system. The notice must include:

• the number, classification or occupation, and work location of the layoffs substantially due to replacement by AI or other automated technology;
• the job functions the affected workers performed that will be automated;
• the category or type of AI or automating technology involved; and
• at the top of the notice, the statement “This notice is for a technology displacement.”

The EDD will publish summaries of these notices and a quarterly statewide summary, so an AI-driven layoff becomes a public record. (A separate EDD reporting provision, which expires January 1, 2029, imposes no duties on employers.)

The hard part is the trigger. The statute doesn't define “in substantial part.” If you're consolidating a call center because a new AI phone system handles most inbound volume, that's an easy case. If you're cutting a team for budget reasons and you also happen to have bought an AI tool, it's harder. Write down the actual business reasons for any reduction at the time you decide on it. If AI is part of the reason, assume it needs to be disclosed. Calling it something else is a bad bet when the EDD is publishing the summaries.

AB 1883: no emotion inference, no neural data

AB 1883 bars an employer from using an AI-based workplace surveillance tool to recognize, infer, or predict an employee's emotional state, or to collect neural data. “Workplace surveillance tool” is broad: it means any system, application, or device that collects employee data, activities, or behaviors by means other than direct observation by a person. The prohibition does not stop you from using a surveillance tool “to ensure safety.”

The practical exposure is in call-center analytics and coaching software that advertise “sentiment” or “mood” scoring of employees. If a vendor's feature list includes emotion detection, turn it off for employee monitoring or get the vendor to confirm in writing that it isn't running. Violations carry a penalty of up to $500 each, enforced by the Labor Commissioner or a public prosecutor, and civil actions can seek injunctive relief, punitive damages, and attorney's fees.

July 1, 2027: the No Robo Bosses Act

SB 947 is the law most employers will need to build a process around, and the six-month runway is there for that reason.

What counts as an “ADS”

An automated decision system is “any computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence” that issues simplified output, such as a score, classification, or recommendation, used to assist or replace human discretionary decisionmaking and that materially impacts people. Spam filters, firewalls, calculators, and similar tools are excluded.

Productivity scores, attendance-pattern flags, and customer-rating thresholds that trigger a write-up can all qualify.

The core rules

• No sole reliance. “An employer shall not rely solely on an ADS when making a disciplinary or termination decision.”
• Primary reliance requires human corroboration. If you primarily rely on ADS output, a human must corroborate the decision using other relevant information, such as supervisory evaluations, personnel files, work product, or peer reviews.
• If it can't be corroborated, you can't use it. If the reviewer can't corroborate the output, or finds it “inaccurate, incomplete, or misleading,” the output can't support the decision.
• Post-use notice. When you primarily relied on an ADS, you must give the employee a written, plain-language notice at the time you tell them about the decision. It must say an ADS was primarily relied upon, that a human reviewed and corroborated the output, how to reach a human for more information (including the employee's right to a description of their data), and that retaliation is prohibited.
• Data access. Employees may request a “meaningful, objective description” of their own data the ADS used, with other people's personal information anonymized.
• Flat prohibitions. Regardless of reliance, you may not use an ADS to violate labor, employment, or civil rights laws, to infer an employee's protected status, or to predict and take adverse action against someone for exercising legal rights.

Enforcement and the burden shift

The Labor Commissioner and public prosecutors enforce SB 947. The civil penalty is $500 per violation, and remedies in a civil action include injunctive relief, punitive damages, and attorney's fees. Commentators note that earlier drafts gave employees an express right to sue and that the final text dropped it. The enforcement section still refers to civil actions generally, so expect plaintiffs' lawyers to test how far that reaches.

The provision to plan around is the burden shift. Once someone shows an ADS was used in a disciplinary decision, the employer has to prove either that it did not primarily rely on the ADS or that it complied with the corroboration and notice rules. That makes it a records problem as much as a legal one.

Two narrow exits exist. A collective bargaining agreement can waive the law, but only “in clear and unambiguous terms” and with protections against algorithmic management. And uses required by federal law or federal contracts in aviation, defense, and similar fields are carved out.

The rules that were already here

SB 947 sits on top of two California regimes that already apply, and complying with one doesn't satisfy the others.

• FEHA automated-decision regulations. The Civil Rights Council's regulations on automated-decision systems took effect October 1, 2025. They apply FEHA's discrimination rules to AI tools used in hiring, promotion, and other employment decisions, and they require keeping employment records, including automated-decision data, for at least four years. FEHA's discrimination provisions cover employers with five or more employees. Our article on AI hiring tools and EEOC exposure covers the federal side, and vendor and agent liability explains why “the vendor did it” isn't a defense.
• CCPA automated decisionmaking technology rules. The California Privacy Protection Agency's regulations were approved September 22, 2025 and took effect January 1, 2026. Their ADMT requirements (pre-use notice, opt-out, and access rights where ADMT is used for “significant decisions,” including hiring, promotion, demotion, suspension, and termination) generally require compliance by January 1, 2027. They apply only to businesses that meet the CCPA's thresholds, most commonly annual gross revenue above $26,625,000 (the inflation-adjusted figure in effect since January 1, 2025, per the CPPA). SB 947 says expressly that CCPA-covered businesses remain subject to the CPPA's ADMT rules. Our state privacy law overview has more.

So a 40-person California company under the CCPA revenue threshold deals with FEHA, SB 947, and AB 1883. A 300-person company over it deals with all of them, plus SB 951 if a layoff is in the cards.

A worked hypothetical

Here's a typical scenario. A 60-employee home-services company in Sacramento dispatches technicians through a field-service platform. The platform scores each technician on job time, callbacks, and customer ratings, and auto-flags anyone below a threshold for three consecutive weeks. The operations manager has been issuing written warnings off those flags and terminated two technicians this year after repeated flags.

After July 1, 2027, the flag-to-warning pipeline is a problem. If the manager is primarily relying on the score, someone has to look at the underlying jobs (were the callbacks the tech's fault or a parts problem?) and document that review before the warning goes out. The employee gets a standalone notice at the time of the warning. If a customer-rating dip turns out to reflect a single angry customer, or a route assignment that gave the tech the longest drives, the output is “incomplete or misleading” and can't be used.

At 60 employees the company is under Cal-WARN's 75 threshold, so SB 951 doesn't apply. FEHA does, so the platform's scoring data falls under the four-year retention rule.

A five-item checklist for a 20–200 person California employer

1. Inventory every tool that scores, ranks, or flags employees. Include features buried inside payroll, scheduling, point-of-sale, and field-service platforms. Ask each vendor whether the feature uses machine learning, statistical modeling, or AI. Under SB 947 the answer is usually yes.
2. Map which tools feed discipline or termination. For each, decide now whether managers will rely on it primarily, partially, or not at all. Write that down in your employee AI use policy.
3. Build the corroboration step and the notice before July 1, 2027. Create a short reviewer form showing what was examined, and a template post-use notice with a named human contact. Fold both into your termination process.
4. Audit monitoring features for emotion or neural data before January 1, 2027. Turn off sentiment scoring of employees, and get vendor confirmation in writing.
5. If you're at or near 75 employees at any California site, update your WARN template. Add the SB 951 fields and set up a decision memo practice for reductions that involve automation.

Frequently Asked Questions

This article reflects California law as enacted through September 30, 2026, and is current as of October 2026. Regulations, agency guidance, and early enforcement may change how these laws are applied. This is general information, not legal advice for your situation.

This article is for educational purposes only and does not constitute legal, tax, or financial advice. Consult a qualified professional for guidance specific to your situation.