California’s No Robo Bosses Act Requires Human Review

California AI workplace law: Governor Gavin Newsom signed a package of worker protections on September 30, 2026. The most consequential measure for employers using algorithmic performance systems is Senate Bill 947, commonly called the No Robo Bosses Act. It bars an employer from relying solely on an automated decision system to discipline or terminate an employee. Where the system is the primary basis for a decision, a human must corroborate its output and the affected employee must receive a written notice. Crucially, these SB 947 duties become operative July 1, 2027; signing the bill did not make them immediately enforceable.

The second major measure, SB 951, adds AI displacement details to California’s existing mass-layoff notices. It does not create a new universal notice for every job changed by AI, and the final bill retains the existing 60-day Cal/WARN notice period. Together, the laws make two different questions harder for employers to avoid: who verified the machine’s conclusion about an individual worker, and what technology is replacing work in a large layoff?

That distinction matters beyond California. AI vendors serving multinational employers, including companies with engineering and operations teams in India, may face more detailed procurement questions from California customers. But the California statute does not automatically govern an employee in India merely because their company buys the same software. Its immediate legal scope is the covered California workplace.

What happened on September 30?

Newsom’s office announced that he signed SB 947, SB 951 and two workplace-surveillance bills, AB 1331 and AB 1883, among a wider set of AI measures. The governor described the package as a move to keep people involved in consequential employment choices and to make technological job displacement more visible. The official announcement is the primary record for the signing; the enrolled SB 947 text and enrolled SB 951 text control the detailed obligations.

Original reporting by Associated Press, CalMatters and Bloomberg Law independently covered the signing. Their reports emphasize different parts of the story: AP describes the broader package, CalMatters examines labor’s partial victory and Bloomberg Law focuses on the new human-review requirement. That is more useful corroboration than counting syndicated versions of a single wire story as separate sources.

The governor’s shorthand that AI cannot decide a firing by itself is broadly right, but it is not the entire rule. SB 947 regulates an “automated decision system,” a category that reaches machine learning, statistical modeling, data analytics and AI processes whose scores or recommendations materially affect people. Ordinary calculators, datasets and some security tools are explicitly excluded. The law also covers discipline, not just termination.

How the No Robo Bosses Act changes an individual decision

SB 947’s central command is clear: an employer cannot rely solely on an automated decision system when deciding to discipline or terminate an employee. If the employer primarily relies on an automated output, it must direct a human to corroborate the conclusion. The statute offers examples of evidence a reviewer may consult, including managerial evaluations, personnel records, the employee’s work product, peer reviews and witness interviews. A manager merely clicking “approve” without examining supporting information would be a poor reading of a rule expressly requiring corroboration.

The bill then sets a hard limit. If the output cannot be corroborated, or the human reviewer concludes that it is inaccurate, incomplete or misleading, the employer cannot use that output for the disciplinary or termination decision. That creates a meaningful separation between using a model as a lead and treating its score as an unquestioned verdict. It does not ban all workplace analytics or require a person to ignore every algorithmic insight.

For a covered decision where the automated system was the primary basis, the employee must receive a separate written notice when informed of the outcome. It must explain that the system was principally used, that a human reviewed and corroborated the result, whom to contact for more information and that retaliation for exercising the law’s rights is prohibited. The employee may ask for a meaningful, objective description of their own data used by the system. The statute simultaneously requires protection of other people’s personal information in that description.

The law also bars an employer from using an automated system to infer an employee’s protected status or to predict and punish workers for exercising legal rights. Those restrictions address risks upstream of a firing decision: a tool could affect opportunities, monitoring or discipline long before a final termination letter appears. They do not make every prediction about workplace performance unlawful; the exact prohibited use and context matter.

Enforcement is narrower than some campaign language suggests. The Labor Commissioner and public prosecutors can act, and the enrolled text provides a $500 civil penalty per violation. CalMatters reported that the final SB 947 lost a worker appeals process and an individual right to sue to compel compliance during legislative negotiations. It also no longer directly covers independent contractors as workers, although the definition of employer includes certain labor contractors. Those differences should be visible in any business assessment of the law.

There are additional limits. A qualifying collective bargaining agreement can expressly waive the statute if it also supplies specified protections against algorithmic management. A narrow exception addresses automated systems needed to comply with federal requirements or binding contracts in specified aviation, national-security, military, space or defense operations. These are written exceptions, not an invitation to label every technology vendor a defense contractor.

AI layoffs are a different problem, governed by SB 951

SB 951 amends the existing California Worker Adjustment and Retraining Notification framework. That framework generally concerns a covered establishment with at least 75 people and a mass layoff of 50 or more employees within a 30-day period, or certain relocation or closure events. The enrolled bill preserves a 60-day advance notice requirement for covered events. Earlier proposals or commentary describing a new 90-day period do not reflect the enrolled version.

When a covered layoff, relocation or termination is caused in whole or substantial part by AI or other automated technology replacing or automating jobs, the employer’s existing notice must disclose more. It must state how many roles are affected, their occupations and locations, which job functions are being automated, and the specific category or type of technology substantially causing the displacement. The notice must carry the phrase “This notice is for a technology displacement” at its top. The California Employment Development Department must publish summaries of these notices and quarterly statewide technology-displacement summaries.

This is an attribution rule inside an existing mass-layoff process. It does not mean that every individual job loss involving a chatbot triggers Cal/WARN, nor that firms must file a separate AI report whenever they halt hiring. SB 951 does require the state agency to report to the Legislature on AI’s effects on business hiring by January 1, 2028. Its definition of a “technological cessation in hiring” supports that study, while recommendations about a possible future notice rule remain recommendations.

By making employers describe the actual work replaced, the law could improve public evidence about AI’s labor-market effects. It could also expose disagreements. A company may attribute a restructuring to lower demand, outsourcing, software automation or some combination. The requirement turns that causal claim into a disclosed piece of a covered notice; it does not itself prove how many jobs AI has displaced across the economy. Researchers should treat later state summaries as reported employer data and examine their scope.

What the accompanying surveillance laws add

The September 30 package was wider than firing and layoffs. The governor’s announcement says AB 1331 restricts workplace surveillance in bathrooms, and AB 1883 limits the use of tools to infer a worker’s emotional state or collect brain data. AP and CalMatters separately reported these measures as part of the same signing. Those provisions deserve their own technical and legal review; they should not be confused with SB 947’s human-review requirement or SB 951’s layoff disclosure.

The difference matters for employers purchasing a suite of products from one vendor. A dashboard might contain productivity scores, video monitoring, sentiment analysis and HR recommendations. Each function can raise a distinct question. Does a score materially influence discipline? Is there a human checking the underlying evidence? Does a sensor infer emotional or brain states? Where can a surveillance device operate? A generic assurance that “a human is in the loop” does not answer all four.

CalMatters’ reporting also shows why the legislative outcome should not be inflated into a comprehensive ban on algorithmic management. Labor groups sought broader measures, and several provisions were reduced or left out. The signed laws place specific guardrails around particular decisions and disclosures while leaving substantial room for AI use elsewhere in the workplace. Their practical force will depend on employer implementation, state enforcement and how the final statutory terms are interpreted.

Why Indian technology and business readers should pay attention

California is a major market for enterprise software and an operating base for many AI developers. An India-based startup selling HR analytics to a California employer may therefore encounter customer requirements for explainability, evidence retention, human-review workflows and notice support. That is a commercial inference from the statute’s employer duties, not a claim that the California Labor Code directly regulates every supplier worldwide.

The product question is concrete. If a customer uses a model’s risk score to recommend terminating an employee, can the tool show which employee data drove that result, preserve a review trail and allow a manager to record corroborating evidence? If the customer cannot explain the score without exposing another person’s private data, the product needs a better disclosure design. Vendors that build these capabilities into normal workflows may face less friction than those offering only a black-box ranking.

For Indian employers operating in California, the relevant analysis is more direct: identify which California employees and establishments are covered and map where automated recommendations enter discipline, termination and large restructuring decisions. A multinational should not assume that a single global HR policy captures local legal requirements. Nor should it assume that a California rule automatically applies to its India workforce. Our broader guide to AI regulation and India’s approach explains why jurisdiction and use case matter.

The law also offers a governance contrast. The recently signed White House AI safety accord sets out voluntary company-level review commitments for advanced AI. SB 947 instead specifies a state-law rule for a particular employment decision, while SB 951 collects information about covered technology-driven layoffs. Neither framework substitutes for the other: a model developer’s safety pledge does not itself verify whether an employer fairly reviewed one worker’s case.

One useful implementation test is to run a simulated decision before July 2027. Give a reviewer a flawed score and ask whether they can locate its underlying evidence, spot the error, decline to rely on the output and generate a compliant plain-language notice if the output was the primary basis. Such a test may reveal more than a policy statement promising “responsible AI.” It also points to an operational cost: review takes staff time, documentation and training. The bill does not prescribe a particular software architecture, so employers have choices about how to meet the standard.

What remains uncertain

Signing fixes the enrolled text but not every future application. The boundary between a system that merely assists and one on which an employer “primarily” relies may be contested. Employers may dispute whether a human’s review genuinely corroborated an output. In mass layoffs, deciding whether technology was a substantial cause can require a counterfactual judgement about what would have happened without automation. The state’s published notice data will eventually make some claims visible, but it cannot remove all ambiguity.

It is also too early to attach a numerical claim about jobs saved, layoffs disclosed or compliance costs to these new statutes. SB 947 has not yet reached its July 1, 2027 operative date, and SB 951’s state hiring report is due later. Vendor demonstrations, union concerns and political claims are useful signals of what stakeholders expect, not measured outcomes of a law that has just been signed.

For now, the soundest reading is precise: California has moved from a debate about “robot bosses” to two defined obligations. An automated score cannot be the only basis for disciplining or firing a covered worker once SB 947 is operative, and a human must check the evidence when it is the primary basis. A qualifying mass-layoff notice must identify AI or other automation when it substantially displaces jobs. The strongest business response is to design for an accountable decision and an honest disclosure, while watching how implementation and enforcement develop.

Frequently asked questions

Does California ban employers from using AI at work?

No. SB 947 restricts specific uses of automated decision systems in discipline and termination and requires human corroboration when their output is the primary basis. Other uses are not automatically banned by that bill.

When does the No Robo Bosses Act take effect?

The enrolled SB 947 text says its relevant workplace provisions become operative on July 1, 2027. Newsom signed it on September 30, 2026.

Does SB 951 require 90 days of notice for AI layoffs?

No. The enrolled measure retains the existing 60-day Cal/WARN notice period for covered events and adds technology-displacement information when AI or other automation is a substantial cause.

Does the law apply to Indian workers?

It is California employment law. Its requirements may shape what multinational buyers expect from Indian vendors, but the law does not automatically govern an employee in India. Specific cross-border situations require jurisdiction-specific analysis.

Source note: Event date: September 30, 2026. Primary sources: California governor’s signing announcement, enrolled SB 947 and enrolled SB 951. Independent original reporting: AP, CalMatters and Bloomberg Law. This report distinguishes signed law from later operative and reporting dates.