In last week’s final scorecard on the Governor’s signing decisions, I listed SB 947, the “No Robo Bosses Act of 2026,” among the technology bills that became law. It needs a fuller treatment. Governor Newsom signed it on September 30. The law adds Part 5.5.5 to the Labor Code (§§ 1520–1526.7) and makes California the first state to require human oversight when employers use automated tools to discipline or fire employees.
This is the Legislature’s second attempt. Last October the Governor vetoed SB 7, which he called overly broad. SB 947 is narrower in four ways:
- It covers employees only. Independent contractors and applicants are not covered.
- It covers only discipline and termination decisions.
- It drops the advance notice that an automated tool is in use.
- It drops the limit on using customer ratings as a system’s primary input.
The name sounds like a ban, but it is not one. Employers can keep using scoring and analytics tools. What the law regulates is the point at which a system’s score becomes an employment decision. The law is not operative until July 1, 2027, so there is time to prepare. The catch is that many employers already use tools that meet the statute’s definition without thinking of them as “AI.” Here are five things employers need to know now.
1. The tool can recommend, but a human has to decide and check the work against the data.
The core rule is in Labor Code § 1522, and it has two levels.
- No sole reliance. An employer may not rely solely on an automated decision system (ADS) to make a disciplinary or termination decision.
- Primary reliance requires human corroboration. If an employer primarily relies on an ADS output to discipline or terminate, it must have a human corroborate the decision. The human must use the data that was collected or used to produce the output, or other relevant information. The statute gives examples: supervisor or manager evaluations, personnel files, work product, peer reviews, and witness interviews. The witness-interview example expressly includes relevant online customer reviews. Restaurant and hotel employers should note that one.
The corroboration requirement has real consequences. If the employer cannot corroborate the output, it may not use the output to discipline or terminate. The same is true if the human reviewer concludes the output is inaccurate, incomplete, or misleading (§ 1522(c)). A manager clicking “approve” on a system-generated recommendation will not meet this standard. The review has to be real, and it has to be documented.
One open question: the statute does not define “primarily rely.” Until there is guidance, employers should adopt an internal standard and apply it consistently. That means deciding which tools you will treat as triggering the law and which outputs are only background information, and writing down the reasoning.
2. “Automated decision system” covers far more than tools marketed as AI.
The statute defines an ADS broadly. It is any computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that produces a simplified output, such as a score, classification, or recommendation. That output must be used to assist or replace human discretionary decision-making that materially affects people. The law excludes basic infrastructure: spam filters, firewalls, antivirus software, identity and access management tools, calculators, databases, and datasets.
The line falls between the data and the score. A database that stores attendance records is excluded. A dashboard that turns those same records into a “reliability” score on a manager’s screen is not. For restaurant and hospitality employers, consider whether any of these tools feed into discipline:
- Timekeeping or scheduling platforms that automatically assign attendance “points” or flag tardiness
- Speed-of-service or productivity scoring (ticket times, drive-thru times, rooms cleaned per shift)
- Loss-prevention analytics that flag voids, comps, discounts, or cash-handling patterns at the POS
- Guest survey or review tools that score individual employees
- Camera or monitoring analytics that generate alerts about employee conduct
If a system’s output is the reason a manager starts or finishes a disciplinary conversation, the system belongs on your inventory.
Two coverage points also matter. First, there is no headcount threshold, and labor contractors and public employers are covered. Second, the law reaches only decisions about existing employees. Hiring tools are outside SB 947, but they are not outside the Civil Rights Council’s FEHA regulations on automated decision systems, discussed in point five.
3. The notice is due at the moment of decision, and it has to match everything else in the file.
Under Labor Code § 1524, an employer that primarily relies on an ADS to discipline or terminate must give the employee a written notice at the time it informs the employee of the decision. The notice must be a separate, standalone communication. It must be in plain language and in the language the employer routinely uses with the employee. It must:
- State that the employer primarily relied on an ADS to make the decision
- Confirm that a human reviewed and corroborated the ADS output
- Provide contact information for a human who can answer questions about the decision
- Explain the employee’s right to request a description of the employee data the ADS used
- State that retaliation for exercising rights under the law is prohibited
That data request is the second piece employers need to plan for. Under § 1522(d), the employee may request a “meaningful, objective description” of the employee’s own data used by the ADS, and the employer must provide it. The description must also anonymize personal information about customers, coworkers, and other individuals (§ 1522(e)). That is harder than it sounds. A loss-prevention flag built on register activity, or a productivity score that ranks employees against each other, mixes the employee’s data with other people’s. Someone has to be able to separate the two.
This notice will become one of the most important documents in any later lawsuit. As I have written before, the reason for a termination has to be consistent across every document. That includes the termination letter, the Notice to Employee as to Change in Relationship, the EDD response, and eventually deposition testimony (see the key issues to address before terminating an employee). An SB 947 notice that tells a different story than the termination letter is a gift to plaintiff’s counsel. Draft a template now and add it to your termination checklist.
One more point: complying with SB 947’s notice requirements satisfies substantially similar notice provisions under other state law (§ 1526.3). That relief does not extend to the California Privacy Protection Agency’s automated decision-making technology regulations for businesses covered by the CCPA (§ 1526.4).
4. Once ADS use is shown, the burden shifts to the employer, and the statute never says how long to keep the proof.
The enforcement scheme is built around a burden shift. Under § 1526.1(c), once it is shown that an ADS was used in a disciplinary decision, the employer must prove one of two things. Either it did not primarily rely on the ADS, or it complied with the corroboration and notice requirements. The employee only has to show that a system was used. Everything after that is the employer’s to prove, with whatever records it kept at the time.
The Labor Commissioner and public prosecutors enforce the law. Violations carry a $500 civil penalty per violation, and § 1526.1(d) makes injunctive relief, punitive damages, and attorney’s fees available. The enacted law dropped the express private right of action that appeared in earlier drafts. However, the burden-shifting provision refers broadly to “any civil action or administrative proceeding brought pursuant to this part.” Expect plaintiffs’ firms to use that language to argue for a broader right to sue. Also expect them to test whether SB 947 violations can be pursued through PAGA.
Three uses are prohibited outright, and no amount of human review cures them (§ 1522(a)). An employer may not use an ADS:
- To violate labor, safety, employment, or civil rights laws
- To infer an employee’s protected status under Government Code § 12940
- To predict that a worker will exercise a legal right and take adverse action on that basis
A tool that labels employees as “likely to complain” or “organizing risks” is a serious liability.
The practical problem that commentary has largely missed is records retention. Corroboration only works if the data behind the output still exists. If the raw inputs behind a six-month-old performance score have been deleted, the employer cannot corroborate that score and therefore cannot use it. And if the decision is challenged, the employer bears the burden of proving what its reviewer actually looked at. SB 947 does not set a retention period. It simply penalizes employers who do not keep the records. This is the same lesson I discussed about payroll records being your first witness: the records testify about the period before you knew you had a problem. Set a retention policy for ADS inputs, outputs, and the reviewer’s corroboration file. At a minimum, it should match how long you keep personnel files, which we recommend be at least four years after separation.
5. Use the runway: a checklist for July 1, 2027.
SB 947 is one layer of a growing California framework. The Civil Rights Council’s FEHA regulations on automated decision systems took effect October 1, 2025. They already make discrimination through automated tools actionable, including in hiring and promotion decisions that SB 947 does not reach. The CPPA’s automated decision-making technology regulations add pre-use notice, opt-out, and access obligations for significant employment decisions beginning January 1, 2027. On the same date, SB 951 requires new disclosures in Cal-WARN notices when AI or automation drives a mass layoff.
Practical steps before July 1, 2027:
- Inventory your tools. Identify every system that generates scores, flags, or recommendations about employees, and map which ones feed into discipline or termination.
- Set your “primary reliance” standard. Decide which tools trigger the law, document why, and apply the standard consistently across locations.
- Ask vendors hard questions, and put the answers in the contract. What data does the tool use, and how long is it logged? Can it reconstruct how a past output was produced? Can it support a description of one employee’s data while anonymizing everyone else’s? SB 947 places the obligations on the employer, not the vendor.
- Write a human review protocol. Specify who reviews ADS-driven discipline, what they must examine, and how they document the corroboration, including a decision not to proceed. A checkbox saying the reviewer “agreed” will not survive the burden shift.
- Build the notice and data-request process. Prepare a template notice in every language you use for routine communications, identify the contact person by role rather than by name so the template survives turnover, and set up a workflow to respond to data-description requests.
- Fix retention. Make sure ADS inputs, outputs, and review files are kept long enough to defend a decision.
- Train managers. The biggest risk is a manager who treats a system alert as the decision. The output is where an investigation starts, not where it ends.
- Review collective bargaining agreements. A CBA can waive the law only if it does so in clear and unambiguous terms, expressly provides for wages and working conditions, and provides protection from algorithmic management (§ 1526.5). Local ordinances with equal or greater protections are not preempted.
The underlying principle is not new. Good discipline and termination decisions have always required a real investigation and good documentation. SB 947 makes that a statutory requirement whenever technology drives the decision, and it puts the burden on the employer to prove the investigation happened.
Want to go deeper on SB 947 and everything else changing for California employers in 2027? Join Zaller Law Group on Thursday, October 15, 2026 at 10:00 AM Pacific for our masterclass, “New Laws and PAGA Update for 2027.” In one hour, we’ll walk through the key employment bills signed this year and their effective dates, the latest PAGA developments two years after the 2024 reforms (including cure and penalty-reduction opportunities), and the handbook, training, notice, and payroll updates to put on your 2027 compliance checklist before January 1. The program is designed for HR professionals, in-house counsel, business owners, and managers, and offers 1.0 hour of California MCLE credit and 1.0 SHRM PDC for attendees who attend the full hour. Register here.









