California will prohibit employers from relying solely on automated decision systems to discipline or fire workers from July 1, 2027, under SB 947, a bill by Senator Jerry McNerney that Governor Gavin Newsom approved on September 30, 2026, and that the Secretary of State chaptered the same day as Chapter 859, Statutes of 2026.

In Short

California passed a law saying a computer program cannot be the only thing that decides whether a worker gets disciplined or fired. If a company mostly relies on such a program, a human has to check the result against other evidence and the worker has to be told, in writing, that software was involved. The rules start on July 1, 2027, and companies that break them can be investigated by the state Labor Commissioner and fined $500 per violation.

What the law requires

SB 947 adds a new Part 5.5.5 to Division 2 of the California Labor Code, under the heading "Automated Decision Systems in the Workplace," beginning at Section 1520. The text comes from the chaptered version dated September 30, 2026, as published on Digital Democracy, the legislative tracking site run by CalMatters.

The definitions set the reach of everything that follows. An automated decision system, or ADS, is defined as "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 decisionmaking and materially impacts natural persons." The statute then removes a list of everyday tools from that definition: "a spam email filter, firewall, antivirus software, identity and access management tools, calculator, database, dataset, or other compilation of data."

ADS output is defined broadly as "any information, data, assumptions, predictions, scoring, recommendations, decisions, or conclusions generated by an ADS." Artificial intelligence is defined as "an engineered or machine-based system that varies in its level of autonomy and that can, for explicit or implicit objectives, infer from the input it receives how to generate outputs that can influence physical or virtual environments."

The definition of employer is wide. It covers anyone who "directly or indirectly, or through an agent or any other person, employs or exercises control over the wages, hours, or working conditions of any person." State government branches, cities, counties, special districts, transit districts, school and community college districts, and the California State University are all included, with the University of California covered only "upon agreement by the regents." Labor contractors, farm labor contractors and foreign labor contractors are pulled in as well.

Three outright prohibitions

Section 1522(a) lists three uses that are banned regardless of how heavily an employer relies on the system. An employer may not use an ADS to "prevent compliance with or violate" federal, state or local labor, health and safety, employment or civil rights laws. It may not use one to infer an employee's protected status under Section 12940 of the Government Code, which is California's fair employment statute. And it may not use one to "predict and take adverse action against a worker for exercising their legal rights."

These three prohibitions are not limited to discipline and termination. On the plain text, they apply to any employer use of an ADS, which would include scheduling, performance scoring and hiring tools, although the rest of the statute is built around disciplinary and termination decisions.

The human corroboration rule

The core of the law sits in Section 1522(b). "An employer shall not rely solely on an ADS when making a disciplinary or termination decision," it states. Where an employer "primarily relies" on ADS output for such a decision, it must "direct a human to corroborate the decision using data that was collected or used to produce the ADS output or other relevant corroborating or supporting information."

The statute names five kinds of supporting material that may be used: supervisory or managerial evaluations, personnel files, employees' work product, peer reviews, and "Witness interviews, that may include relevant online customer reviews." The list is introduced with "may include, but is not limited to," so it is illustrative rather than closed.

Section 1522(c) closes the loop. If an employer cannot corroborate the output, or if the human reviewer concludes the output is "inaccurate, incomplete, or misleading," the employer may not use it to make the disciplinary or termination decision at all.

Two thresholds are therefore at work. "Solely" relying on an ADS is forbidden outright. "Primarily" relying on one is permitted, but only with human corroboration, a notice, and a data description on request. The law does not define "primarily," which leaves a gap that the Labor Commissioner or the courts will presumably have to fill.

Data description and post-use notice

Under Section 1522(d), an employee may request "a meaningful, objective description" of their own data used by the ADS when an employer has primarily used one to make a disciplinary or termination decision, and the employer must provide it. Section 1522(e) requires that description to anonymize the personal information of customers, other employees or other individuals.

Section 1524 adds a written post-use notice. An employer that primarily relied on an ADS for a disciplinary or termination decision must give the affected employee that notice "at the time the employer informs the employee of the decision." It must be written "in plain language as a separate, stand-alone communication," in the language normally used for workplace communications, and delivered through "a simple and easy-to-use method, including an email, hyperlink, or other written format."

The notice must state four things: that the employer primarily relied on an ADS; that a human reviewed the decision and corroborated the output; contact details for a human the employee can ask about the decision and about the right to the data description; and that the employer cannot retaliate against the employee for exercising rights under the new part.

Enforcement, penalties and the burden of proof

Section 1526 prohibits retaliation. An employer may not discharge, threaten, demote, suspend or "in any manner discriminate or retaliate" against an employee for using the rights, filing a complaint with the Labor Commissioner, alleging a violation, or cooperating in an investigation.

The Labor Commissioner is the primary enforcer under Section 1526.1(a). The office may investigate, issue citations, file civil actions and order "appropriate temporary relief to mitigate a violation or maintain the status quo" while a full investigation or hearing is pending. Section 1526.1(b) allows public prosecutors to enforce the part as well.

The fine is modest. Section 1526.1(e) sets "a civil penalty of five hundred dollars ($500) per violation." Section 1526.1(d) goes further for court actions brought in the county where the violation occurred or where the person resides or does business, stating that the petitioner "may seek appropriate temporary or preliminary injunctive relief, including punitive damages, and reasonable attorneys fees and costs." The pairing of a preliminary injunction with punitive damages in one clause is unusual drafting, since punitive damages are normally a final remedy rather than an interim one. The text does not resolve that tension.

The most consequential procedural provision may be Section 1526.1(c), which shifts the burden of proof. In any civil action or administrative proceeding over Section 1522(b) or (c), "once it has been demonstrated that an ADS was used to make a disciplinary or deactivation decision, the employer must demonstrate that the employer did not primarily rely upon an ADS or that the employer complied with Sections 1522 and 1524." A worker therefore needs to show only that a system was used. The employer then has to prove either that its reliance fell short of "primarily" or that it followed the corroboration and notice steps.

Where the law steps aside

Several provisions limit the statute's reach or its interaction with other rules.

  • Local ordinances. Section 1526.2 states that the part does not preempt any city or county ordinance that "provides equal or greater protection."
  • Notice safe harbor. Section 1526.3 says an employer that complies with the notice requirements is not required to comply with "any substantially similar notice provisions" related to ADS in employment under other state law. That safe harbor does not cover quotas, as defined in Labor Code Section 2100, or "any other automated standards applied to working conditions."
  • Privacy regulator carve-back. Section 1526.4 overrides the safe harbor for businesses subject to the California Consumer Privacy Act of 2018. Those employers remain subject to "any privacy-related automated decisionmaking technology regulation duly adopted by the California Privacy Protection Agency."
  • Collective bargaining. Section 1526.5 exempts parties covered by a collective bargaining agreement only if it waives the part "in clear and unambiguous terms," provides for wages and working conditions, and "provides protection from algorithmic management."
  • Federal work. Section 1526.6 excludes ADS use required by, or reasonably necessary to comply with, federal statutes, regulations or binding contracts relating to aircraft development for the national airspace, or to "national security, military, space, or defense purposes."
  • Charter cities. Section 2 of the act declares the matter one of statewide concern, so the law applies to all cities, including charter cities.

Section 1526.7 sets the operative date: "This part shall become operative on July 1, 2027."

What the law does not contain

The chaptered text places every obligation on employers. There is no requirement for ADS developers or vendors to supply documentation, no impact assessment, no bias audit, no registration and no pre-use notice to workers before a system is deployed. Hiring decisions receive no corroboration or notice rule; those provisions are written only for "disciplinary or termination" decisions.

That shape is narrower than McNerney's previous attempt. SB 7, which his office called the "No Robo Bosses Act," cleared the Assembly 45-17 and the Senate 28-9 in September 2025, according to a September 12, 2025 release from McNerney's Senate office. According to a client alert from the law firm Kilpatrick Townsend, Newsom vetoed SB 7 on October 13, 2025, calling its restrictions overly broad and raising concern that they could stop companies from using technology to identify and reward high-performing workers, for instance by factoring in customer ratings. The same alert describes SB 7 as requiring advance notice to workers before such systems were used.

SB 947 drops the advance notice and keeps only the post-use notice. It also names "relevant online customer reviews" as acceptable corroborating evidence, which reads as a direct answer to the customer ratings concern, though the attached pages do not say so and the link is an inference.

How the bill moved

The Status tab on Digital Democracy lists the full legislative history. The bill was introduced on February 2, 2026, and could be acted on from March 5. It was amended on March 26 and referred to the Senate committees on Labor, Public Employment and Retirement, and on Privacy, Digital Technologies, and Consumer Protection.

Recorded votes, as listed in the Status tab:

  • April 8, 2026 - Senate Labor, Public Employment and Retirement: 3 ayes, 1 no.
  • April 20, 2026 - Senate Privacy, Digital Technologies, and Consumer Protection: 7 ayes, 2 noes.
  • May 4, 2026 - placed on the Senate Appropriations suspense file.
  • May 14, 2026 - Senate Appropriations, do pass as amended: 5 ayes, 2 noes.
  • May 19, 2026 - Senate floor: 29 ayes, 9 noes.
  • June 10, 2026 - Assembly Labor and Employment: 5 ayes, 1 no.
  • June 23, 2026 - Assembly Privacy and Consumer Protection: 10 ayes, 4 noes.
  • June 30, 2026 - Assembly Judiciary: 8 ayes, 3 noes.
  • August 13, 2026 - Assembly Appropriations: 10 ayes, 4 noes.
  • August 21, 2026 - read a third time and amended on the Assembly floor; Assembly Rule 69(b)(1) suspended, 55 ayes, 17 noes.
  • August 30, 2026 - Assembly floor: 53 ayes, 14 noes.
  • August 31, 2026 - Senate concurrence in Assembly amendments: 28 ayes, 10 noes.

The bill was enrolled and presented to the Governor "at 2 p.m." on September 9, 2026. Newsom approved it on September 30, the same day it was chaptered. The Assembly passage on August 30 and the Senate concurrence on August 31 came at the very end of the session, after the August 21 floor amendments that produced the final text.

Ash Kalra, Eloise Gómez Reyes and Chris Ward are listed as co-authors for the chaptered version. Gómez Reyes, a senator, is also the author of SB 1130, a separate bill on secret recordings by smart glasses in workplaces, which PPC Land covered in February 2026.

Discrepancies between the tabs

The four pages do not agree on every number, and two of the differences matter for anyone citing the vote.

The Votes tab lists the Senate concurrence as 29 for, 10 against and 1 not voting, dated August 30, 2026. The Status tab records the same action on August 31 as "Ayes 28. Noes 10." The Votes tab total of 40 matches the full Senate; the Status figure of 38 does not. The pages do not explain the gap. One possible explanation, which is inference rather than anything stated in the documents, is a member adding a vote after the roll call closed.

Every committee and floor date in the Votes tab also sits one day earlier than the matching entry in the Status tab and in the "Discussed in Hearing" list. Assembly Appropriations appears as August 12 in one and August 13 in the other; Senate Labor appears as April 7 and April 8. The pattern is consistent across all eleven roll calls, which suggests a time zone or data entry offset in the Votes tab rather than different events. This article uses the Status tab dates.

The Votes tab adds non-voting counts that the Status tab omits: 2 not voting on the May Senate floor vote, 12 on the August 30 Assembly floor vote, and 1 in each of the four Assembly committees.

The statute text itself has one inconsistency. Section 1526.1(c) shifts the burden of proof once an ADS is shown to have been used to make "a disciplinary or deactivation decision," while every other operative section refers to "disciplinary or termination" decisions. "Deactivation" is the term usually applied to app-based workers. The law does not define it, and its application to people the law does not treat as employees is unclear.

The Analysis tab, which normally holds committee analyses, was not among the pages provided, so legislative staff estimates of cost or scope are not available here.

Who lined up on each side

The Supporters and Opponents tab draws positions from bill analyses. It lists 140 entries: 45 marked support and 95 marked against. Several organizations appear more than once under slightly different names or dates. The American Petroleum and Convenience Store Association, the Association of California Healthcare Districts, PRISM, the Protect App-Based Drivers and Services Coalition and California's Credit Unions each appear twice, and the SMART Transportation Division legislative board three times. Counting only the most recent entries, dated August 30, 2026, gives 40 supporters and 87 opponents.

Supporters were dominated by organized labor and civil society groups. They include the California Federation of Labor Unions, SEIU California, the California Teachers Association, the California Nurses Association, the California Federation of Teachers, AFSCME California, UAW Region 6, the Communications Workers of America District 9, the National Union of Healthcare Workers and numerous regional labor councils. Technology and privacy groups on the support side include the Electronic Frontier Foundation, Oakland Privacy, TechEquity Collaborative, Tech Oversight California and the Kapor Center Advocacy. The California Employment Lawyers Association and The Greenlining Institute also backed it.

Opposition came from business associations, local governments and technology firms. The California Chamber of Commerce, the California Business Roundtable, the California Retailers Association, the California Restaurant Association, the California Hospital Association, the California Bankers Association and dozens of local chambers are listed against. So are the Business Software Alliance, TechNet, the Silicon Valley Leadership Group and the Chamber of Progress.

Three app-based companies or coalitions are on the opposing side: DoorDash, Lyft, Inc. and Uber Technologies, Inc., alongside the Protect App-Based Drivers and Services Coalition. Staffing firms opposed through the American Staffing Association, the California Staffing and Recruiting Association and California Staffing Professionals. SHRM California and the Society of Human Resources Management are also listed against, as is the Insights Association, which represents the market research industry.

Public employers were split from their unions. The League of California Cities, the California State Association of Counties, the Urban Counties of California, the California School Boards Association, the Association of California School Administrators and the counties of Fresno, Kern and Ventura opposed the bill, while public sector unions including the California School Employees Association and the Orange County Employees Association supported it.

Why it matters for marketing and advertising

Advertising agencies, ad tech companies and publishers are employers like any other, and many of them run the kind of performance dashboards, productivity scoring and quota systems that could qualify as ADS under the new definition. From July 1, 2027, any California disciplinary or termination decision that primarily relies on that output needs human corroboration and a written notice. The burden shift in Section 1526.1(c) means record-keeping on how decisions were reached becomes the defence.

The software vendors that sell these systems are not regulated directly, but their trade bodies, including the Business Software Alliance and TechNet, opposed the bill. Workday already faces a federal age discrimination case: PPC Land reported in August 2025 that a court in the Northern District of California expanded the Mobley v. Workday collective action to include applicants processed through HiredScore AI features. PPC Land's April 2026 coverage of a study on the data that workplace apps collect found Workday was the only one of 10 apps that did not let users request deletion of their data.

The explicit reference to "relevant online customer reviews" as corroborating evidence also reaches into marketing territory. Reviews on third-party platforms become a documented input to employment decisions, which raises the stakes on review accuracy for businesses whose staff are rated publicly.

California's approach differs from the other frameworks PPC Land has followed this year. Colorado's SB 26-189 repealed and replaced the state's 2024 AI law in May 2026, covering employment among several categories of consequential decisions and placing documentation duties on developers as well as deployers, with Attorney General enforcement only. Connecticut's Senate passed SB 5 with provisions on hiring tools in April 2026, including pre-decision notices and explanations after adverse decisions. SB 947 instead sits in the Labor Code, is enforced by a labor regulator and public prosecutors, and governs only the employer.

The federal backdrop is less settled. The White House in March 2026 published a national AI policy framework that seeks to override state AI rules considered burdensome, while preserving state authority over consumer protection. In July 2026, the FTC proposed a policy statement that named Colorado's amended law as potentially preempted. Whether a Labor Code rule aimed at employer conduct, rather than at how AI models are built, would attract the same attention is an open question.

In Europe, workplace AI is regulated through two channels. The GDPR already restricts solely automated decisions with legal or similarly significant effects, and an Austrian court in September 2025 found that a public employment service's algorithm complied with Article 22 because advisers could override it, the same human-in-the-loop logic that SB 947 writes into statute. Under the EU AI Act, employment screening is classed as high-risk, and a provisional agreement on May 7, 2026 pushed those obligations to December 2, 2027, five months after California's rules take effect.

California has been busy on AI legislation more broadly. On September 18, 2026, Newsom signed an executive order setting a November 16 deadline for recommendations on a frontier AI kill switch, alongside bills on AI auditors and companion chatbots signed earlier that month. SB 947 is narrower than any of those. It does not touch how models are trained or sold. It governs what a California employer must do before software output becomes a reprimand or a dismissal.

Timeline

Summary

Who: California Senator Jerry McNerney, as author, with co-authors Ash Kalra, Eloise Gómez Reyes and Chris Ward; Governor Gavin Newsom, who approved the bill; California employers, including public employers; the Labor Commissioner and public prosecutors as enforcers; 40 organizations in support, led by labor unions, and 87 opposed, led by business groups, technology firms and local governments, on the most recent count.

What: SB 947 adds Part 5.5.5 to the California Labor Code. It bars employers from relying solely on automated decision systems for disciplinary or termination decisions, requires human corroboration when they primarily rely on one, requires a written post-use notice and a data description on request, prohibits three uses outright, shifts the burden of proof to employers once ADS use is shown, and sets a $500 civil penalty per violation. It places no duties on vendors and contains no audit or pre-use notice requirement.

When: Approved and chaptered on September 30, 2026, as Chapter 859, Statutes of 2026. The rules become operative on July 1, 2027.

Where: California, applying to all employers in the state, including charter cities, with exemptions for qualifying collective bargaining agreements and certain federal defense and aviation work.

Why: The law follows Newsom's October 2025 veto of McNerney's broader SB 7 and narrows the approach to discipline and termination, dropping advance notice while keeping human review. For agencies, ad tech firms and publishers that use performance scoring or productivity systems, it sets a documented human review step before software output can lead to discipline or dismissal, at a time when federal authorities are questioning state AI rules.