California bans fully automated layoffs and requires human review

Governor Gavin Newsom has signed SB 947, nicknamed the "No Robo Bosses Act," prohibiting employers from delegating disciplinary and termination decisions solely to automated decision systems. The law requires that every such decision be reviewed by a person, and that the employee receive a full explanation of the reasoning — including personnel files, evaluations, work product, peer reviews and witness interviews. If the employer cannot verify the system's output, or the human reviewer finds it wrong, partial or misleading, the output may not be relied on for disciplinary action or termination. The law takes effect July 1, 2027.

The legislation arrives as data show how deeply AI has already penetrated HR processes. A MyPerfectResume survey of 1,000 HR professionals in April found that 73% use AI for hiring decisions, 52% use it for restructuring planning and role design, and 51% use it to flag "risky" candidates. Only 26% of respondents said they do not use AI for layoff decisions, meaning the vast majority already involve an algorithm at some stage of the process.

The law does not ban automated systems outright; it mandates a layer of human oversight and transparency toward the employee. Employers will have to disclose whether mass layoffs, relocations or role closures were driven by an AI system, according to a briefing from the governor's office. An employee who believes the process was not followed can file a complaint with the California Labor Commissioner. The statute leaves room for interpretation, however: it does not define what constitutes "meaningful human review" and imposes no criminal penalty for violations.

Lily Li, an AI-focused attorney and founder of Metaverse Law, told ZDNET the law may set a precedent for other states, but she expects employers to try to circumvent its spirit. Companies will likely perform "checkbox reviews" — formal, low-cost procedures that satisfy the letter of the law without substantively examining the system's output, she said. The employee's right to demand an explanation acts as a partial brake, but the practical burden may fall on the fired workers themselves, who would have to sue to challenge decisions that were not properly reviewed.

At the federal level, regulation still rests on voluntary agreements with leading labs and is considered less concrete. States are taking divergent approaches: Colorado and Connecticut require only disclosure of AI use; Illinois enacted a first-of-its-kind law this summer mandating third-party audits of frontier labs; California and New York are pursuing a stricter line. Until a federal version of these protections is adopted, workers will continue to encounter a patchwork of safeguards that varies from state to state.