California Enacts New Workplace AI Rules: What SB 947 and AB 1883 Require

The laws address different points in the workplace-technology lifecycle. SB 947 focuses on how employers use automated outputs, particularly in discipline and termination. AB 1883 focuses on two sensitive kinds of collection or inference through workplace surveillance tools. Newsom also signed SB 951, a related measure addressing information in Cal/WARN notices when a covered workforce reduction is caused substantially by AI or other automation.

This article describes the final text signed by the Governor and official materials verified on September 30, 2026. It provides general regulatory information, not legal advice. The Secretary of State had not yet posted chapter assignments when this article was published; the substantive text and signing status are final, and chapter citations will be added when officially available.

Status of SB 947, AB 1883, and SB 951

Why these laws matter

California employers already face legal questions when software screens applicants, evaluates performance, recommends discipline, monitors activity, or processes sensitive worker data. The new laws do not create one comprehensive workplace-AI code. Instead, they add targeted Labor Code rules to a layered system that already includes FEHA, the CCPA and its final ADMT regulations, wage-and-hour rules, labor protections, and other privacy obligations.

The practical distinction is important. SB 947 regulates reliance on an automated decision system and attaches procedural safeguards to certain adverse decisions. AB 1883 prohibits two specified uses of AI-enabled surveillance tools regardless of whether the resulting information is used to make an employment decision. Existing discrimination and privacy law continues to apply to the same workflow where its separate elements are met.

What SB 947 requires

• California SB 947 automated discipline and termination recordReview the tracked status, dates, scope, and official primary sources.

SB 947 adds a new Labor Code part governing automated decision systems in the workplace, operative July 1, 2027. Its definition of an automated decision system covers a 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—used to assist or replace human discretionary decision-making and materially affecting natural persons. Listed exclusions include spam filters, firewalls, antivirus software, identity-and-access-management tools, calculators, databases, datasets, and other compilations of data.

The employer definition reaches a person that employs or controls wages, hours, or working conditions, directly, indirectly, through an agent, or through another person. It expressly includes state and local governmental entities, specified education and transit entities, labor contractors, farm labor contractors, and foreign labor contractors. The bill's operative protections are framed around employees rather than applicants or independent contractors.

• An employer may not use an ADS to prevent compliance with or violate labor, occupational-safety, employment, or civil-rights law.
• An employer may not use an ADS to infer an employee's protected status under Government Code section 12940.
• An employer may not use an ADS to predict and take adverse action against a worker for exercising legal rights.
• An employer may not rely solely on an ADS when making a disciplinary or termination decision.
• When an employer primarily relies on ADS output for discipline or termination, it must direct a human to corroborate the decision using the data used to produce the output or other relevant supporting information.
• If the output cannot be corroborated, or the reviewer concludes it is inaccurate, incomplete, or misleading, the employer may not use it to make the disciplinary or termination decision.

The enrolled bill does not establish a general rule for every consequential employment action. Its express no-sole-reliance, corroboration, data-description, and postuse-notice provisions concern disciplinary and termination decisions. Hiring, promotion, compensation, scheduling, and work allocation may remain governed by FEHA, CCPA, or other law, but the enrolled SB 947 text does not attach the same decision-specific procedure to those actions.

SB 947 notice, access, and enforcement

When an employer primarily relies on an ADS for discipline or termination, the affected employee may request a meaningful, objective description of the employee's own data used by the system. The description must anonymize personal information belonging to customers, other employees, or other individuals. The statute does not create a broad right to inspect source code, model weights, or every item of evidence.

At the time it communicates the decision, the employer must also provide a separate, plain-language postuse notice in the language used for routine employee communications. The notice must state that the employer primarily relied on an ADS, that a human reviewed the decision and corroborated the output, how to contact a human for more information and request the data description, and that retaliation is prohibited.

The law authorizes enforcement by the Labor Commissioner and public prosecutors. It specifies temporary relief, citations, civil actions, potential punitive damages, reasonable attorney's fees and costs, and a $500 civil penalty per violation. For specified discipline or termination claims, once use of an ADS is shown, the employer bears the burden of demonstrating that it did not primarily rely on the system or that it complied with the safeguards. The text contains anti-retaliation language but does not state a general employee-initiated private right of action in the same direct terms as its grants to the Labor Commissioner and public prosecutors; this article therefore does not characterize it as an unrestricted standalone private action.

A qualifying collective bargaining agreement could expressly waive the new part if it clearly and unambiguously does so, covers wages or earnings, working conditions, and other terms, and provides protection from algorithmic management. A separate narrow exemption would cover specified aircraft, national-security, military, space, or defense operations only where ADS use is required by or reasonably necessary to comply with a federal statute, regulation, or binding federal contract.

What AB 1883 requires

• California AB 1883 workplace surveillance recordReview the tracked status, dates, scope, and official primary sources.

AB 1883 adds a separate Labor Code part covering workplace surveillance tools. It defines those tools broadly as systems, applications, instruments, or devices that collect or facilitate collection of employee data, activities, communications, actions, biometrics, or behaviors by means other than direct human observation. Examples include video or audio surveillance, continuous incremental time tracking, geolocation, electromagnetic tracking, photoelectronic tracking, and photo-optical systems.

The actual prohibition is much narrower than that definition. An employer may not use a workplace surveillance tool that uses artificial intelligence on employees to recognize, or make inferences or predictions about, an individual's emotional state, or to collect neural data. Neural data means information generated by measuring activity of an employee's central or peripheral nervous system and not inferred from nonneural information.

AB 1883 does not prohibit workplace surveillance as a category. It does not prohibit ordinary productivity monitoring, time tracking, geolocation, video, audio, facial recognition, behavioral monitoring, or biometric collection merely because those functions fit the tool definition. Those practices may implicate other laws, and a covered tool would violate AB 1883 if it performed one of the two prohibited AI-enabled functions, but the enrolled bill should not be summarized as a broad workplace-surveillance ban.

The text states that the prohibition does not prevent use of a workplace surveillance tool to ensure safety, or otherwise, when the tool does not meet the two prohibited specifications. This is best read as confirming the narrow scope of the prohibition, not as a free-standing safety exception authorizing emotion inference or neural-data collection. The bill also includes the same kind of limited federal-compliance exemption for covered aircraft, national-security, military, space, or defense operations.

AB 1883 coverage and enforcement

The employer definition includes private persons and governmental entities that employ or control wages, hours, or working conditions, plus state and local government, the University of California, California State University, specified districts, labor contractors, farm and foreign labor contractors, and entities contracting with a labor contractor for labor or services. The operative prohibition applies to use on employees; the enrolled text does not extend it expressly to job applicants or independent contractors.

The Labor Commissioner may investigate, issue citations, order temporary relief, and file a civil action; public prosecutors may also enforce the law. A civil action may seek temporary or preliminary injunctive relief, punitive damages, and reasonable attorney's fees and costs. A violation can carry up to $500 per violation. The text says an employee, the Labor Commissioner, or a public prosecutor may recover the penalty as a statutory penalty paid to the employee or as a civil penalty, but not both for the same violation.

Newsom signed AB 1883 but asked lawmakers to clarify it

Newsom's September 30 signing message supports guardrails on increasingly sophisticated workplace surveillance and identifies the risk that granular employee information can enable manipulative or abusive practices. At the same time, he wrote that AB 1883 lacks key definitions, which may create confusion about which tools are covered, and encouraged the Legislature to refine the measure to address ambiguity and possible unintended consequences.

That message does not suspend, narrow, or rewrite the enacted text. Until any amendment is enacted, employers and vendors must evaluate coverage against the statute's existing definitions and the actual functions of a tool. Future clarification is possible because the Governor requested it, but this article does not assume that lawmakers will amend the law or predict what any amendment would say.

What changed during the legislative process

The enrolled AB 1883 is materially narrower than the broad workplace-surveillance measure suggested by its title and by descriptions of earlier versions. The final text centers on only two prohibited functions: AI-based emotional-state recognition or inference and neural-data collection. It does not retain a general code governing notice, purpose limitation, productivity metrics, protected activity, or every biometric or behavioral monitoring practice.

That narrowing changes the compliance analysis. Employers should not import removed concepts from older summaries into the final bill. Facial recognition and biometric information remain within the definition of data a surveillance tool may collect, but the final prohibition is not triggered by those categories alone. Likewise, the bill does not create a general ban on monitoring productivity or protected activity. Other California laws may still regulate those practices.

SB 947 also ended as a targeted measure rather than a blanket prohibition on automated employment decisions. It distinguishes sole reliance from primary reliance: sole reliance is prohibited for discipline and termination; primary reliance triggers human corroboration, employee-data access, and postuse notice. Those are different legal tests and should remain separate in policies and training.

Human review and automated employment decisions

Under enrolled SB 947, human involvement is not satisfied merely by placing a manager at the end of an automated workflow. When the employer primarily relies on the output, the human must corroborate the decision using relevant information. If corroboration fails, or the reviewer finds the output inaccurate, incomplete, or misleading, the employer may not use that output for discipline or termination.

The bill does not use the terms appeal or reconsideration and does not grant an express right to reverse the decision through a separate appeal process. The employee receives a postuse notice, a human contact, and a right to request a description of the employee data used. Calling that package an appeal right would overstate the enrolled text.

Workplace surveillance and sensitive inference

AB 1883 draws a statutory line around two uses of particularly sensitive employee information. Emotional-state inference can involve voice, face, movement, physiological signals, or other inputs, but the legal question is whether an AI-enabled workplace surveillance tool is recognizing, inferring, or predicting emotional state. Neural-data coverage is narrower still: the information must come from measurement of central or peripheral nervous-system activity and may not be merely inferred from nonneural information.

The text does not separately prohibit facial recognition or general biometric processing. A camera or biometric system may be a workplace surveillance tool, but AB 1883 applies only if the AI-enabled tool performs one of the prohibited functions. Employers still need a separate review of privacy, discrimination, labor, and sector-specific rules for monitoring that falls outside AB 1883.

How the laws fit with FEHA

California's FEHA framework already applies employment-discrimination rules to automated-decision systems. Civil Rights Council regulations effective since October 1, 2025 make clear that using an automated system or selection criterion can violate FEHA when it causes unlawful discrimination. The rules also address disability-related inquiries and require retention of covered employment records, including defined automated-decision-system data, generally for at least four years.

SB 947 and AB 1883 add targeted Labor Code requirements without replacing that analysis. A manager can corroborate an automated termination recommendation under SB 947 and still face a FEHA issue if the practice discriminates. An emotion-inference system can violate AB 1883 even without proof of discriminatory effect. Compliance with one framework addresses one set of duties, not a safe harbor from the other.

• California AI Hiring Compliance in 2026: How FEHA and CCPA Automated Decision Rules Fit TogetherA detailed guide to the existing discrimination and privacy frameworks.
• California state intelligence pageCurrent California rules, future dates, and source-backed records.
• AI Bias Audits: One-Time Compliance or Continuous Monitoring?How recurring audit requirements differ from broader governance practices.

How the laws fit with CCPA automated-decision rules

The CCPA is a separate privacy regime for covered businesses processing personal information. Final regulations took effect January 1, 2026. Risk assessments already apply to specified new high-risk processing, while Article 11's pre-use notice, access, opt-out, and applicable human-appeal provisions begin January 1, 2027 for covered ADMT used to make significant decisions.

Enrolled SB 947 expressly preserves duly adopted CPPA automated-decisionmaking regulations for an employer that is a CCPA-covered business. Its limited coordination provision for substantially similar state notice duties does not displace CCPA privacy obligations. An SB 947 postuse notice after discipline or termination is also conceptually different from the CCPA's pre-use notice and individual rights.

The definitions also differ. SB 947 reaches systems that assist or replace human discretionary decision-making, while the final CCPA ADMT definition turns on technology that replaces or substantially replaces human decision-making. A workflow may therefore fall within SB 947 even when genuine human involvement keeps it outside Article 11's ADMT definition. Conversely, CCPA risk-assessment duties may be triggered by categories of processing beyond the final SB 947 safeguards.

Practical comparison

How common workplace tools may fit

Employers and vendors have different operational roles

The new laws principally regulate employers within their statutory definitions. An employer can remain responsible for the employment decision even when a vendor supplies the model, score, monitoring device, or interface. Contract language allocating compliance work does not necessarily determine statutory responsibility.

Vendors can nevertheless create operational and compliance risk depending on system design and their actual role. Employers may need vendors to identify prohibited or optional features, explain what employee data drives an output, preserve evidence needed for corroboration, and support notices and access requests. That does not mean every technology vendor is directly liable under SB 947 or AB 1883; coverage turns on statutory role, conduct, and workflow.

What employers should evaluate now

1. Inventory decision workflows. Identify automated tools that influence discipline or termination and document whether outputs are advisory, primary, or effectively dispositive.
2. Test the review process. Confirm that managers can understand the output, examine relevant supporting information, identify errors or omissions, and decline to use an uncorroborated result.
3. Map notice responsibilities. Determine who could produce an SB 947 postuse notice, provide a human contact, and respond to a request for a meaningful description of employee data.
4. Inventory monitoring. Identify video, audio, geolocation, time-tracking, biometric, wearable, neurotechnology, and behavioral-analysis systems used on California employees.
5. Ask what systems infer. Determine whether a product attempts to infer emotion, intent, attention, stress, fatigue, mental state, or related traits and distinguish emotional-state inference from safety functions or other analytics.
6. Trace the data. Record which employee, customer, and third-party information feeds each tool, what outputs are produced, who receives them, and how long records are kept.
7. Separate legal frameworks. For each workflow, evaluate SB 947 or AB 1883 alongside FEHA, CCPA, wage-and-hour, labor, and other applicable rules rather than treating one review as comprehensive.
8. Clarify vendor responsibilities. Contracts should support data access, explanations, notice fields, audit trails, change management, and restrictions on prohibited features, while preserving the employer's ability to make an independent decision.

Related bill: SB 951 and technological displacement

Newsom also signed SB 951 on September 30. Beginning January 1, 2027, it amends Cal/WARN notice content when a covered mass layoff, relocation, or termination is caused in whole or substantial part by an AI system or other automated technology replacing or automating positions. The notice must identify affected job counts, classifications or occupations, locations, automated functions, and the category or type of technology, and state at the top that it is a technology displacement notice.

SB 951 does not regulate individual algorithmic discipline or workplace surveillance. It operates through the existing Cal/WARN framework, including its covered-establishment and mass-layoff thresholds, rather than creating notice for every AI-related staffing change. It also requires the Employment Development Department to publish summaries and submit a report by January 1, 2028 on AI's effects on hiring, including possible future notice requirements for a technological cessation in hiring. That report provision repeals January 1, 2029.

Effective dates and implementation timeline

Key takeaways

• SB 947 is not a blanket ban on automated employment decisions. It prohibits sole ADS reliance for discipline and termination and adds safeguards when reliance is primary.
• SB 947 requires corroboration, not merely nominal human involvement, but it does not create an express employee appeal or reconsideration process.
• AB 1883 is not a general workplace-surveillance ban. Its final prohibition targets AI-enabled emotional-state recognition or inference and neural-data collection.
• Neither bill replaces FEHA discrimination obligations or CCPA privacy and ADMT duties.
• SB 951 is relevant to larger technology-driven workforce reductions, not individual automated decisions.
• Newsom signed SB 947, AB 1883, and SB 951 on September 30, 2026; their substantive requirements begin in 2027.