What California SB 947 Would Mean for AI Discipline
California’s enrolled SB 947 targets automated employee discipline and firing. See its human-review, notice and enforcement provisions.

California’s “No Robo Bosses Act,” SB 947, is not law as of September 25, 2026. It is an enrolled bill awaiting Governor Gavin Newsom’s action. If enacted, it would prohibit employers from relying solely on automated decision systems to discipline or terminate current employees, require genuine human corroboration when automation is the primary basis, and require written notice in those primary-reliance cases. It would become operative July 1, 2027—not immediately.
The official bill record says SB 947 was presented to the governor on September 9, 2026. Under Article IV, Section 10 of the California Constitution, he has until September 30 to sign or veto it; if it is not returned by then, it becomes law without his signature. Employers can prepare now, but should not present SB 947’s proposed safeguards as current law.
Choose how AI was used and whether notice was given; the result applies the enrolled SB 947 framework.
SB 947 Decision Check
Default: AI flagged an employee for termination, no human reviewed it, and no notice was given.
Solely automated termination would violate SB 947
If enacted and operative, SB 947 would prohibit relying solely on an ADS for this decision. No notice can cure the absence of human decision-making.
Enforcement: A worker could bring the matter to the California Labor Commissioner. The enrolled bill authorizes the Labor Commissioner and public prosecutors to enforce it and sets a $500 civil penalty per violation; it does not expressly create a standalone private right of action.
What counts as corroboration and notice?
Corroboration means checking the output against relevant supporting information, such as evaluations, personnel files, work product, peer reviews or witness interviews. When an employer primarily relies on an ADS, the separate written notice must disclose that reliance and human review, identify a human contact, explain the employee-data access right and state the retaliation protection.
Source: enrolled SB 947, proposed Labor Code §§1520–1526.4. The bill was pending on September 25, 2026 and would be operative July 1, 2027 only if enacted.
SB 947 Would Ban Solely Automated Discipline and Firing
SB 947 focuses on automated systems affecting current employees, particularly discipline and termination. It is not a general AI hiring law for applicants.
The enrolled bill defines an automated decision system, or ADS, as a computational process derived from machine learning, statistical modeling, data analytics or AI that produces a simplified output—such as a score, classification or recommendation—used to assist or replace human discretion and materially affect people. It expressly excludes tools such as calculators, databases, datasets, spam filters, firewalls, antivirus software, and identity and access-management tools (enrolled SB 947 text, proposed Labor Code §1520).
The bill would impose three levels of restriction:
- Some uses would be prohibited outright. An employer could not use an ADS to violate labor, safety, employment or civil-rights rules; infer an employee’s protected status; or predict and take adverse action because a worker exercised legal rights.
- An ADS could not make the decision alone. An employer could not rely solely on an ADS when deciding to discipline or terminate an employee.
- Primary reliance would require human corroboration. If an employer primarily relied on an ADS output, a human would have to corroborate it with relevant supporting information. Examples in the bill include manager evaluations, personnel files, employee work product, peer reviews and witness interviews.
If the output could not be corroborated—or the reviewer determined that it was inaccurate, incomplete or misleading—the employer could not use the output for that decision (enrolled SB 947 text, proposed §1522).
Putting a manager’s name on an automated result would not, by itself, satisfy that standard. The contemplated reviewer must test the output against supporting evidence.
Notice Would Follow Decisions Based Primarily on an ADS
The notice obligation would apply when an employer primarily relied on an ADS to discipline or terminate an employee. It would not attach merely because an automated tool played some lesser role.
At the time the employee is told of the decision, the employer would have to provide a separate, plain-language written notice in the language used for routine employee communications. The notice would have to state:
- that the employer primarily relied on an ADS;
- that a human reviewed the decision and corroborated the output;
- contact information for a human who can provide more information;
- the employee’s right to access a description of their own data used by the ADS; and
- that retaliation for exercising rights under the proposed law is prohibited.
The enrolled text does not say that the notice must identify the system by name or reproduce every data point. It gives the employee the right to request a “meaningful, objective description” of their own data used by the ADS, with other people’s personal information anonymized (enrolled SB 947 text, proposed §§1522–1524).
This is narrower than some accounts of earlier versions. Compliance plans should use the enrolled September 4 text rather than the February proposal or an alert describing a prior amendment.
Coverage Extends Beyond Large Private Employers
The proposed definition of employer contains no minimum employee count. It includes private employers, labor contractors, farm and foreign labor contractors, and many state and local public bodies. The bill would apply to charter cities as a matter of statewide concern.
A limited exception would apply where a qualifying collective bargaining agreement clearly waives the law and provides the specified working-condition and algorithmic-management protections. Another narrow exception covers ADS use required by, or reasonably necessary for, certain federal aviation, national-security, military, space or defense obligations.
Local ordinances providing equal or greater employee protection would not be preempted. Businesses covered by the California Consumer Privacy Act would also remain subject to applicable privacy-related automated-decision regulations adopted by the California Privacy Protection Agency.
Enforcement Would Not Create a Standalone Employee Lawsuit
The enrolled bill authorizes enforcement by the California Labor Commissioner and public prosecutors, prohibits retaliation, and sets a civil penalty of $500 per violation. It also permits specified court remedies and attorney’s fees in those enforcement actions (enrolled SB 947 text, proposed §§1526–1526.4).
The final text does not expressly create a standalone private right of action for an employee. Accordingly, the $500 provision should not be described as an automatic personal award or a general right for every affected worker to file a civil claim. That differs from descriptions published earlier in 2026, before later amendments.
For a worker, the practical first step would be to preserve the decision notice, request the description of personal data used by the ADS where the bill provides that right, and document indications that no person genuinely reviewed the result. The enrolled bill’s specific enforcement route is through the Labor Commissioner and public prosecutors, rather than an express standalone employee claim.
SB 947 Is Narrower Than the Vetoed SB 7
SB 947 is a rewritten successor to SB 7, which Newsom vetoed in 2025 as overbroad. The enrolled SB 947 centers on discipline and termination decisions involving current employees. Its operative rules distinguish among sole reliance, primary reliance and lesser use of an ADS.
That structure matters. The bill does not ban all workplace AI or every automated contribution to a personnel decision. It bars an ADS from being the sole decision-maker, requires human corroboration when the system is the primary basis, and connects the post-use notice to that primary reliance.
The enrolled version also lacks an express standalone private right of action. Earlier summaries of the proposal may therefore overstate either its scope or the remedies available to an individual worker. A September 2026 legislative update also describes the measure as part of the Legislature’s later package of AI employment bills.
HR Teams Should Map Evidence, Not Just Systems
An inventory should identify systems that generate performance, productivity, attendance, safety, conduct or risk scores, then record how each output reaches a disciplinary or termination workflow. For each system, determine:
- whether the output can trigger discipline or termination;
- whether it is the sole basis, primary basis or only one lesser input;
- which independent records a human can use to corroborate it; and
- whether the employer can retrieve and objectively describe the affected employee’s input data.
The record should capture the reviewer’s evidence and conclusion, not only an approval click. If the reviewer finds an output inaccurate, incomplete, misleading or unsupported, the proposed rule would prevent its use in the decision.
Employers would also need a process for delivering a separate notice at the same time as the disciplinary or termination decision when an ADS was the primary basis. That process must provide a human contact and support the employee’s request for a description of their own data.
Applicant-screening systems should be tracked separately. SB 947’s enrolled text concerns employees, while anti-discrimination, privacy and local automated-employment rules may separately govern hiring. For broader California context, see whether an employer can fire someone using AI.
The Bill’s Status Must Be Checked Before Compliance Begins
SB 947 could be signed, vetoed or allowed to become law without a signature by September 30, 2026. If enacted, it would become operative July 1, 2027. Confirm the official bill status and any chaptered text before treating that date as a binding compliance deadline. This is general information, not legal advice.