California Enacts Ban on AI-Only Firing Decisions
The new laws also require disclosure of technology-driven mass layoffs and add duties for healthcare providers and AI developers.
California Gov. Gavin Newsom (D) signed legislation on September 30 prohibiting employers from relying solely on automated systems to fire or discipline workers. When an employer relies primarily on an automated recommendation to fire or discipline a worker, a human must review employee records or other relevant information to confirm that it supports the decision. Its requirements begin July 1, 2027.
The latest workplace laws are part of a 13-law package that also restricts employee surveillance, protects clinicians’ judgment when using AI, and expands requirements for identifying AI-generated media.
Employers must check automated firing decisions
Senate Bill 947, introduced by state Sen. Jerry McNerney (D-Pleasanton), applies to systems that use AI or other data analysis to recommend or make employment decisions. When an employer primarily relies on a system to discipline or fire someone, the employer must assign someone to check the recommendation. An employer cannot use a recommendation it cannot substantiate, or that the person assigned to verify it finds inaccurate, incomplete, or misleading.
Employers must give affected workers written notice explaining the AI system’s role in the employment decision and confirming that a human reviewed the decision. Workers can request details of what data was used in the decision. The law prohibits retaliation for exercising those rights.
The Labor Commissioner and public prosecutors can enforce SB 947 against employers, with a $500 civil penalty per violation. Exemptions include union agreements that expressly waive the law, and specified federal aviation and defense requirements.
Employers must disclose automation behind mass layoffs
SB 951 adds information to California’s existing 60-day notice requirement for qualifying mass layoffs, relocations, and business closures. If AI or other automation causes all or a substantial part of the job displacement, employers must identify the affected roles, the work being automated, and the type of technology responsible.
California’s existing mass-layoff notification law generally covers facilities with at least 75 employees. It defines a mass layoff as at least 50 employees losing their jobs within 30 days.
Separately, AB 1883 prohibits employers from using AI surveillance to infer employees’ emotions or collect data measuring nervous-system activity, subject to limited federal compliance exemptions.
AB 1331 restricts workplace bathroom surveillance, with specified exceptions.
Healthcare providers must preserve clinicians’ judgment
AB 1979 requires hospitals, clinics, and physicians’ offices to take reasonable steps to keep licensed healthcare professionals free to make their own decisions when using AI recommendations about a patient’s diagnosis, treatment, or timing of care. Those organizations cannot use AI to independently perform clinical work that legally requires a professional license. Licensing boards can seek court orders against violations that amount to practicing without a license.
SB 503 assigns duties to developers and healthcare organizations that use AI to help decide when or how to diagnose or treat patients. Both must make reasonable efforts to identify systems at risk of bias. Developers must also make reasonable efforts to reduce that risk and provide the healthcare organizations using the system with information about its risks and how it was tested. Healthcare organizations must regularly monitor affected systems and take reasonable steps to reduce biased effects on patients’ care.
More AI providers must identify generated media
SB 1000 expands the California AI Transparency Act by removing its previous threshold of more than one million monthly users. The requirements now reach smaller providers of publicly accessible generative AI systems in California, subject to exemptions, including a temporary exemption for disability assistive technology.
Providers must offer a free way to check whether they created or altered images, audio, or video beyond minor edits. Where technically feasible, providers must embed information identifying the system and whether it created or altered the content. SB 1000 took effect immediately upon signing.
AB 2713 revises existing editing duties for large online platforms, including making information about media’s origin accessible to users and, where technically feasible, not stripping that identifying information from content that users upload, share, or download. These duties concern information using widely adopted technical standards and begin January 1, 2027.
New measures build on California’s earlier AI laws
Newsom signed SB 53 in September 2025, requiring large developers of the most powerful AI models to publish their safety plans and follow them. The law also requires developers whose models meet its threshold for computing power used in training to report specified serious safety incidents to the state.
California enacted two laws governing outside AI assessors in September. SB 813 requires the state to decide by January 2028 what expertise and independence organizations must demonstrate to receive state recognition for assessing potential harm from AI systems. AB 1405 requires auditors checking whether AI systems comply with California law to register with the state’s Government Operations Agency starting in January 2029. Neither law requires companies to hire an assessor. Businesses can choose to hire these assessors; the two laws regulate the assessors rather than require companies to commission audits.
The media transparency rules began in 2024 with SB 942, which required larger AI providers to embed identifying information in AI-generated media and offer free detection tools.
The new healthcare, layoff disclosure, and surveillance laws take effect on January 1, 2027. SB 947’s requirements for automated discipline and firing decisions start on July 1, 2027.
