SB 947 would require human review before AI fires or disciplines California workers
What SB 947 requires employers to do before using AI in discipline or termination decisions, and what happens if Newsom doesn't sign by September 30.
California's SB 947, the No Robo Bosses Act, awaits Governor Newsom's decision by September 30, 2026. The bill would restrict how California employers use automated decision systems—computational processes derived from machine learning, statistical modeling, data analytics, or artificial intelligence—when making discipline and termination decisions.
If Newsom signs the bill, it takes effect July 1, 2027. If he does not act by September 30, the bill automatically becomes law under California's rule that unsigned bills passed before September 1 become law without a signature. If enacted, SB 947 would be the first state law in the nation to impose these specific guardrails on AI in employment decisions, narrowly focused on discipline and termination rather than the broader approach Newsom rejected in 2025.
What counts as an automated decision system under the law
SB 947 defines an automated decision system as any computational process derived from machine learning, statistical modeling, data analytics, or artificial intelligence that produces simplified outputs used to assist or replace human judgment in ways that materially affect individuals. The definition is intentionally broad to capture algorithms, scoring systems, and predictive models. It distinguishes between high-risk systems that directly affect workers and lower-risk administrative tools by focusing on systems whose outputs materially affect employment decisions.
This definition matters because different rules apply depending on how employers use the system. A scheduling algorithm used to optimize shift assignments might be an ADS, but SB 947's core restrictions apply only to systems used primarily for discipline or termination. The law's scope reflects concerns about high-stakes employment decisions where algorithmic error or bias can most directly harm workers.
Prohibited uses and three categorical restrictions
SB 947 establishes three categorical prohibitions that apply to all ADS use in employment, regardless of whether human review occurs afterward. First, employers cannot use ADS in ways that would violate labor, occupational health, employment, or civil rights laws. Second, systems cannot infer employees' protected characteristics under California's Fair Employment and Housing Act—including race, gender, religion, disability, or other protected statuses. Third, ADS cannot predict or enable adverse actions against workers exercising legal rights.
Beyond these categorical prohibitions, SB 947 adds specific restrictions tied to particular employment decisions. Employers cannot use ADS to conduct predictive behavior analysis—meaning they cannot predict or infer a worker's behavior, beliefs, intentions, personality, or emotional state. Employers also cannot use worker data to inform compensation decisions unless they can clearly demonstrate that any differences in pay for substantially similar work are based on cost differentials in performing the task, or that the data was directly related to the tasks the worker was hired to perform. These provisions target abuse patterns such as algorithms that flag workers for discipline based on inferred attitude or personality, and opaque pay systems that claim cost justification without evidence.
Human review and corroboration requirements
The centerpiece of SB 947 is its human review mandate. Employers cannot rely solely on an automated decision system to make discipline or termination decisions. When an employer primarily relies on ADS output for these decisions, it must direct a human to corroborate the decision using the data collected by or used to produce the ADS output, or other relevant corroborating information. Critically, the human reviewer cannot proceed if corroboration fails. The law requires independent verification—a second look at the facts using the same underlying data—before any adverse action can be taken.
This requirement goes beyond simply having a person review the ADS recommendation. The person must independently evaluate whether the recommendation is supported by the underlying data. If an ADS recommends terminating an employee based on low productivity scores, the human reviewer must separately confirm that the productivity data is accurate and that the scores justify termination. If corroboration reveals the data was inaccurate or the recommendation unsupported, the employer cannot proceed with the termination. This creates a meaningful check on algorithmic error.
Notice, transparency and worker rights
When employers primarily rely on ADS for discipline or termination decisions, they must provide affected employees with timely written notice in plain language before, during, or immediately after the decision. This notice must state that an ADS primarily influenced the decision and that a human reviewed and corroborated the output. The notice must include contact information for the human reviewer so workers can ask questions about how the decision was made. Employers must also explain how workers can access a meaningful, objective description of their own data used by the system.
SB 947 creates an affirmative worker right to data access. Workers may request copies of the data descriptions related to their ADS decision once per year. The employer must provide this information in a format the worker can understand—not buried in technical jargon or raw data exports, but as a meaningful account of what information the system used and how it was used. The law also includes strong anti-retaliation protections: employers cannot discharge, threaten, demote, suspend, or discriminate against any employee for taking actions to assert their rights under the law, such as requesting their data or questioning an ADS-driven decision.
“When an employer primarily relies on automated decision system output for discipline or termination, a human must independently verify the decision using the underlying data; if corroboration fails, the employer cannot proceed with the adverse action.”
Enforcement and penalties
SB 947 allows enforcement through two channels: the California Labor Commissioner and public prosecutors, such as the state attorney general or local prosecutors. The bill does not create a private right of action, so workers cannot sue their employers directly for violations.
Violations carry a $500 civil penalty per breach, which can accumulate quickly if an employer makes multiple ADS-driven decisions without proper human review or notice. In a civil action brought by the Labor Commissioner or a public prosecutor, courts may also award punitive damages and attorney's fees on top of the base penalty. The Labor Commissioner can investigate complaints and pursue remedies such as reinstatement for wrongfully terminated workers and back pay. An employer facing a single violation could owe $500 in civil penalties to the state, and if the Labor Commissioner or a public prosecutor brings a civil action and prevails, the employer may also owe punitive damages, attorney's fees, and reinstatement or back pay.
History: Why this version differs from the vetoed SB 7
SB 947 is a reintroduced version of SB 7, which Governor Newsom vetoed in October 2025. Newsom's veto letter criticized SB 7 for imposing overly broad notification requirements on businesses using even routine digital tools like scheduling software, failing to distinguish between high-risk algorithmic systems and low-risk administrative technologies. He also noted that SB 7 overlapped with California's new Automated Decisionmaking Technology regulations finalized by the state Privacy Protection Agency, which already regulate automated workplace decision-making in some contexts. Industry groups warned SB 7 could cost employers over $1 billion in compliance.
SB 947 addresses these concerns by narrowing its focus. Rather than applying broadly to any automated decision system affecting workers, SB 947 concentrates on discipline and termination decisions—the contexts where algorithmic error or bias poses the greatest risk. The bill excludes routine administrative systems and focuses enforcement mechanisms while maintaining the core requirement that humans must independently verify major employment decisions. The narrower scope reflects a compromise between employer concerns about compliance burden and worker advocates' push for algorithmic accountability in high-stakes decisions.
Timeline and what happens if the bill becomes law
SB 947 passed the California Legislature on August 30-31, 2026, reaching Governor Newsom's desk on September 4, 2026. Under California law, bills passed before September 1 that the Governor does not act on by September 30 automatically become law. If Newsom does not sign or veto by September 30, SB 947 becomes law without his signature.
If SB 947 becomes law, employers have nine months until July 1, 2027, to implement compliant systems and processes. Companies using automated decision systems in discipline or termination will need to audit their existing practices, document their human review procedures, revise their notice templates, and prepare systems to provide workers with required data access.
Related coverage: What California's AI transparency rules require of businesses; SB 813 and AB 1405 speed up California's AI auditor rules after Newsom's order.
