What's new in California: A closer look at recently enacted Senate Bill 947—“No Robo Bosses" Act

On September 30, 2026, Governor Gavin Newsom signed into law Senate Bill (SB) 947, the “No Robo Bosses” Act. The law arrives only one year after Newsom vetoed a previous version of the bill as too broad and unnecessarily burdensome on California businesses. SB 947, however, suffers from the same defects. 

Proponents characterize SB 947 as a modest measure that requires employers who rely on artificial intelligence (A.I.) to make disciplinary or termination decisions to verify the A.I.’s output before acting. In reality, the law imposes obligations in several situations beyond discipline and termination, and it creates another basis for litigation under the Private Attorneys General Act (PAGA).

This bill was sponsored by the California Federation of Labor Unions, AFL-CIO. The organization’s president, former Assemblywoman Lorena Gonzalez, has been candid about the broader policy objectives behind the legislation. In her remarks at the “A.I. and Future of Work” global dialogue in Rome on May 5, 2026, Gonzalez presented the future as a choice between two worlds: one in which workers have “security in good jobs, robust protections, and the ability to collectively govern their lives and workplaces,” and one in which workers are displaced by unbridled capitalism. The “best way” to achieve the former is achieved, she observed, “is through engaging workers in the collective bargaining process. But absent universal unions, we need regulations.” This bill is just one such regulation. Employers in California should be prepared for additional regulations in this area in the coming legislative sessions.

SB 947’s key provisions

SB 947 adds new sections to the California Labor Code to regulate the use of an “automated decision system” (ADS). An ADS is broadly defined to include “any computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that issues simplified output, including a score, classification, or recommendation, that is used to assist or replace human discretionary decision-making and materially impacts natural persons.”  

SB 947 prohibits an employer from relying solely on an ADS to discipline or terminate an employee. For employers who “primarily” rely on an ADS output, a human must corroborate the output by reviewing the underlying data or other relevant information. But if the output cannot be corroborated, or is otherwise inaccurate, incomplete, or misleading, the employer may not rely on it to discipline or terminate an employee. The bill also prohibits employers from using an ADS to predict an employee’s future behavior and then taking adverse action based on that prediction.

In addition to corroborating an ADS output on which it primarily relies, the employer must notify the employee in writing, plainly stating that: (1) the employer primarily relied on an ADS output, (2) a human conducted an independent investigation and corroborated the output, and (3) the employee may request a description of the data analyzed by the ADS. If a description is requested, the employer must furnish one that is “meaningful” and “objective.” The legislation also prohibits any employer from retaliating against an employee for requesting a copy of the ADS output relied upon in the personnel management decision.

The legislation authorizes the Labor Commissioner to enforce its requirements and permits public prosecutors, such as city attorneys, to initiate civil or administrative actions. Once it has been shown that the employer “primarily used” an ADS to make a disciplinary or termination decision, the employer “must demonstrate” either that (1) it did not primarily rely on the ADS, or that (2) it corroborated the ADS output and gave proper notice to the employee. A $500 civil penalty attaches to each violation of SB 947. In civil actions, SB 947 also authorizes injunctive relief, punitive damages, and reasonable attorneys’ fees and costs.

Notably, the legislation does not apply to parties covered by a valid collective bargaining agreement if the agreement expressly waives the legislation’s provisions in clear and unambiguous terms and provides protection from algorithmic management.

The PAGA quagmire

While SB 947 does not authorize a private right of action, plaintiffs’ attorneys will likely seek to recover the $500 civil penalty per violation under PAGA. PAGA authorizes aggrieved employees to sue their employers for violations of Labor Code sections that provide for civil penalties recoverable only by the Labor Commissioner “[n]otwithstanding any other provision of law.” And because SB 947 writes its requirements into the Labor Code, they may fall within PAGA’s reach. In other words, even though the fact that SB 947 has no private right of action, employees will likely still bring PAGA actions seeking the $500 penalty. 

As explained above, SB 947 includes a burden-shifting provision that will complicate litigation under this statute. Once a plaintiff shows that the employer used an ADS, then the employer must “demonstrate” either that it did not primarily rely on the system or that it complied with the corroboration and notice requirements. This framework will make defending these claims burdensome for two reasons.

First, the plaintiffs’ bar may argue that the statute shifts not just the burden of proof—as in employment discrimination cases—but the burden of persuasion. Unlike Title VII burden-shifting, in which the plaintiff retains the ultimate burden to prove pretext, SB 947’s requirement that employers “demonstrate” compliance may be used to argue that employers must prove no violation of the statute occurred.

Second, if the employer primarily relied on an ADS, the mandatory post-use notice to the employee supplies the very evidence needed to shift the burden, since the plaintiff need only prove use of an ADS in the initial showing.

Moreover, because the employer must demonstrate compliance, plaintiffs will likely pursue discovery into, at a minimum (1) the data analyzed by the ADS and the resulting output; (2) the ADS’s underlying algorithms and methodology—much of which may be proprietary and commercially sensitive; and (3) other information reviewed by the human corroborator.

Altogether, this legislation will likely force employers to face PAGA claims that necessitate extensive discovery, and to bear an onerous burden that may require litigation through summary judgment to establish compliance. 

Where we stand and what lies ahead

As we move forward, employers should thoroughly review their current processes and consider implementing the following practices: 

1. Inventory your ADS tools now. The statute defines an “automated decision system” broadly. It reaches any machine learning, statistical modeling, data analytics, or A.I. process that produces a score, classification, or recommendation used to assist or replace human decision-making. Employers should identify every tool currently used that feeds into discipline or termination decisions, including productivity scoring and performance analytics. Many of these tools may not be marketed as “A.I.,” so employers must be thorough in this review to ensure all relevant tools are properly considered and cataloged. 
2. Build a documented human corroboration process. Importantly, SB 947 shifts the burden to employers to “demonstrate” compliance.  Given this burden, contemporaneous records will be the best defense. Before acting on an ADS output on which it primarily relies, the employer should have a trained reviewer independently check the underlying data and other relevant information. That review and its conclusions should be documented. If the output cannot be corroborated, or is inaccurate, incomplete or misleading, the employer should not rely on it. Employers should also ensure that no adverse action rests on an ADS prediction of an employee’s future behavior.
3. Standardize the required written notice and data description. Employers should prepare template notices that plainly state three things: that the employer primarily relied on an ADS output, that a human independently investigated and corroborated it, and that the employee may request a description of the data that the ADS analyzed. Employers should also set up a documented process for producing a “meaningful” and “objective” description when one is requested. As part of their annual training, managers should be trained on the anti-retaliation protections for employees who request a copy of the ADS output as part of their annual training requirements.
4. Assess collective bargaining implications. Unionized employers should review whether their collective bargaining agreements meet the requirements for the exemption. Those requirements include an explicit waiver and protections against algorithmic management, which may not be in current collective bargaining agreements. Non-union employers should expect the exemption to give unions added leverage in organizing campaigns.
5. Expect more to come. The legislation’s sponsors have publicly called SB 947 “a good start” and urged a stronger version next year. Employers should monitor upcoming legislation, which may revive provisions dropped from earlier drafts, such as coverage of independent contractors and “deactivation” decisions.

If you have any questions regarding this legislation or how to best implement changes to your company’s A.I. policies and practices, please reach out to the authors of this article.