California employers will have to do more than approve an automated recommendation before disciplining or firing a worker under SB 947, the No Robo Bosses Act. Signed by Governor Gavin Newsom on September 30, the law requires human corroboration when employers primarily rely on automated decision systems, alongside written disclosures and employee rights to a description of the data used. Those workplace requirements become operative July 1, 2027.
The law does not prohibit automated tools from contributing to disciplinary or termination decisions. Instead, it bars employers from relying exclusively on them and establishes an evidence-checking requirement when their output is the primary basis for a decision. It applies to California employers in both the public and private sectors, regardless of size, and protects employees rather than independent contractors.
When the corroboration requirement applies, an employer must assign a human to check the decision against either the data underlying the automated output or other relevant supporting information. That information can include supervisory evaluations, personnel records, employee work product, peer reviews, witness interviews or relevant online customer reviews. The statute permits either route; it does not require reviewers to use every listed source.
The review must have consequences. If the output cannot be corroborated, or the reviewer determines that it is inaccurate, incomplete or misleading, the employer cannot use it to make the disciplinary or termination decision. Simply accepting a system’s score or recommendation without checking supporting evidence does not meet that requirement.
One implementation question remains unresolved in the statutory language: what constitutes primarily relying on an automated system. The law uses that threshold without defining it. Its definition of an automated decision system is broader than generative AI, covering computational processes based on machine learning, statistical modeling, data analytics or artificial intelligence that produce scores, classifications or recommendations used to assist or replace consequential human judgment. Basic tools such as calculators, databases, spam filters and antivirus software are excluded.
Employees must receive a separate written notice when they are informed of a disciplinary or termination decision that primarily relied on such a system. The document must use plain language and the language ordinarily used to communicate with that employee at work. It must disclose the primary reliance on automation, confirm human review and corroboration, identify a human contact, and explain the employee’s data-description rights and protection against retaliation.
On request, the employer must supply a meaningful, objective description of the employee’s own data used by the system in connection with the decision. Employee data includes information that identifies, relates to or describes the worker, whether collected directly, inferred or otherwise obtained. The description must anonymize personal information belonging to customers, other employees or other individuals. This is a right to a description of the relevant employee data, not a statutory promise of access to a year’s worth of records.
Separate restrictions apply regardless of how heavily an employer relies on automation. Employers cannot use these systems to infer an employee’s protected status, evade applicable labor or civil rights obligations, or take adverse action because a worker exercises legal rights—or because a system predicts the worker will do so. The law also prohibits retaliation for requesting its protections, filing a complaint or assisting an investigation or enforcement action.
Enforcement powers include Labor Commissioner investigations, citations, temporary relief and civil actions, as well as enforcement by public prosecutors. The law provides for penalties of $500 per violation, with injunctive relief, punitive damages and attorney fees also available. Before July 1, 2027, employers using automated disciplinary tools face the practical task of establishing a corroboration process, preparing the required notices and ensuring they can describe the employee data behind a decision.
Sources: Atkinson, Andelson, Loya, Ruud & Romo, Crowell & Moring LLP, Epstein Becker Green, sd05.senate.ca.gov, cio.com, Morgan Lewis