NON-SATIRICAL REPORT · Field Notes

California Requires Human Review of AI-Backed Discipline. What Counts as Review?

California has enacted two workplace AI laws that create different future obligations. SB 947 addresses AI-backed discipline and termination beginning July 1, 2027. SB 951 adds disclosures to certain Cal-WARN notices beginning January 1, 2027. The public test is not whether a person appeared in the workflow, but what evidence shows that person actually assessed the automated output.

By Field Notes Editorial Desk · October 27, 2026

Law and date coverage is current through October 27, 2026. Both measures are enacted, but their relevant duties have future operative dates. This report does not establish how any employer will implement the statutes or whether a future workforce event will qualify for Cal-WARN notice.

A human name in the workflow is not the point. The reporting question is whether the record shows a human judgment based on evidence beyond the score.

Two enacted laws, two future dates

California enacted SB 947 and SB 951 on September 30, 2026. They concern different workplace moments and do not begin on the same timetable. SB 951, which changes the contents of certain notices under California's Worker Adjustment and Retraining Notification Act, or Cal-WARN, takes effect January 1, 2027 under the regular effective-date rule for this non-urgency statute. SB 947's rules governing automated decision systems in discipline and termination become operative July 1, 2027. Neither date makes the other law operative sooner. [1] [2] [3]

That sequence matters for readers trying to follow a fast-moving policy debate. In 2027, qualifying Cal-WARN notices may provide an early public record of workforce reductions tied in whole or substantial part to AI or other automated technology. Several months later, SB 947's narrower rules for individual discipline and termination decisions are scheduled to begin. The statutes are enacted, but this report does not assume that employers already follow either future duty or that any particular employer will do so. [1] [2]

SB 947 draws two lines around automated discipline

SB 947 applies to discipline and termination decisions. It first says an employer may not rely solely on an automated decision system, or ADS, to make one of those decisions. It then establishes specific duties when the employer primarily relies on ADS output in making the decision. The distinction is important. Sole reliance is prohibited. Primary reliance is the separate trigger for human corroboration and worker notice. Reporting should not collapse those two phrases into one undifferentiated standard. [1]

Where primary reliance is present, the statute directs the employer to have a human reviewer corroborate the ADS output with data used by the system or other relevant supporting information. The law identifies possible sources of support, including managerial evaluations, personnel files, work product, peer reviews, and witness interviews. If the output cannot be corroborated, or if the reviewer finds it inaccurate, incomplete, or misleading, the employer may not use the output for the discipline or termination decision. [1]

The statute also requires a separate written notice when an employer tells a worker about a discipline or termination decision primarily based on an ADS. The notice must state that the employer primarily relied on an ADS and that a human reviewed and corroborated its output. It must give the worker a human contact, describe the right to seek a meaningful, objective description of the worker's own data used by the system, and state the law's anti-retaliation protection. The chaptered text assigns enforcement roles to the Labor Commissioner and public prosecutors and provides for a civil penalty of $500 per violation. [1]

What a meaningful review is, and is not

SB 947 does not prescribe a minimum review time, reviewer training, independence from the original decisionmaker, or a particular documentation form. It also does not create a general human-review rule for hiring, promotion, scheduling, or every other workplace decision. Those limits should be retained when the law is described. [1]

They do not make the law's corroboration requirement empty. A click confirming an automated score may show that a person was present in a workflow, but it does not by itself answer whether the person corroborated that output with the system's data or other relevant supporting information. That is a reporting question about evidence and process, not an extra mandate added to the statute. Once SB 947 is operative, useful scrutiny will ask what information a reviewer considered, whether the reviewer could reject or change the output, and what the required worker notice says. A generic assertion that a human was involved will not resolve those factual questions. [1]

This is also not a finding about present employer conduct. The law's future operation leaves no basis here to claim that any review is a rubber stamp, that any specific ADS is inaccurate, or that a violation has occurred. The practical public-interest question is narrower: when a consequential decision follows an algorithmic recommendation, can the available record show a human judgment grounded in substantive evidence?

SB 951 adds disclosure to a limited Cal-WARN universe

SB 951 does not create a notice duty for every layoff or every automation project. It changes the information required in notices that are already covered by Cal-WARN. Under the framework retained in the chaptered law, a covered establishment has employed at least 75 people in the preceding 12 months. A mass layoff is a layoff of 50 or more employees during a 30-day period, and the statute's existing notice framework generally calls for 60 days' notice before a covered mass layoff, relocation, or termination. Those thresholds and event definitions continue to limit the law's reach. [2]

For a qualifying Cal-WARN event, the added disclosure applies when displacement by AI or another automated technology caused the mass layoff, relocation, or termination in whole or substantial part. The notice must carry the statement, "This notice is for a technology displacement," at its top. It must also identify the number, classification or occupation, and work location of affected layoffs; the job functions to be automated; and the category or type of system involved. [2]

The statute further directs the Employment Development Department to publish summaries of those notices and quarterly statewide summaries. This could make a bounded public record of qualifying technology-displacement events more legible. It does not convert a company announcement, a hiring freeze, or a smaller or otherwise uncovered job reduction into a Cal-WARN event. Nor does it prove that an employer's stated technology connection is the only cause of a covered reduction. [2]

The documents worth watching

The first date to watch is January 1, 2027, when SB 951 takes effect. The relevant question is whether a covered Cal-WARN notice meeting the law's causation condition contains the required technology-displacement statement and details. There may be no immediate notice, because the duty depends on a qualifying event. [2] [3]

The second date is July 1, 2027, when SB 947's discipline and termination rules become operative. Its core promise is not a ban on workplace AI. It is a demand that an employer cannot rely solely on an ADS for those decisions, and that primary reliance carries corroboration and notice duties. The quality of the resulting review will be a question for records, affected workers, employers, and public enforcement, not a conclusion that can be assumed in advance. [1]

Source: California Legislative Information: SB 947 chaptered text