The short answer
California now requires a human in the loop before AI can discipline or fire anyone, and bans AI that reads workers’ emotions. AB 1883 applies from January 1, 2027 and SB 947 from July 1, 2027. Teams building enterprise HR and workforce systems need human-review workflows, worker notices and emotion-free monitoring before California customers deploy.
What did California sign?
The package landed on Newsom’s final bill-signing deadline of his term. “AI should expand opportunity — not come at the expense of workers,” he said, according to the Associated Press, adding that states have to act “in the absence of federal leadership.” Three bills matter most for software teams.
SB 947, the No Robo Bosses Act, stops employers from relying solely on an automated decision system to discipline or terminate a worker. A human reviewer must independently corroborate the system’s output with other information — manager evaluations, peer reviews, personnel files — and the employee must be told after the fact that an automated system was used. The law also bars using such systems to infer protected characteristics or to predict and retaliate against workers who exercise legal rights.
AB 1883 targets workplace surveillance tools — video and audio monitoring, time trackers, geolocation and similar systems that collect data other than by direct human observation. Employers may not use AI in these tools to recognize or predict emotional states or to collect neural data, though safety uses remain allowed. Bloomberg Government reports penalties of up to $500 per violation, enforced by the labor commissioner and public prosecutors. SB 951 updates Cal/WARN so mass-layoff notices disclose when AI substantially caused the job losses.
When do the rules apply?
- September 30, 2026 — bills signed.
- January 1, 2027 — AB 1883 emotion-recognition and neural-data ban takes effect.
- July 1, 2027 — SB 947 human-corroboration and notice duties take effect.
Newsom also asked lawmakers to clarify the scope of AB 1883 and a companion restroom-surveillance bill, so expect clean-up amendments in 2027. Build to the signed text now and track the follow-up.
Which software is affected?
The duties fall on employers, but the features that create the risk live in software. That puts these product categories in the blast radius: HRIS and performance-management suites that score or rank employees; workforce-analytics tools that flag “low performers” or attrition risk; call-center and sales-coaching platforms that run sentiment or emotion analysis on agents; productivity and bossware tools with webcam, keystroke or voice analytics; and gig and logistics platforms that auto-deactivate workers based on metrics.
If a product can move a worker toward discipline or termination without a documented human decision, or infers mood, stress or engagement from video, voice or biometrics, California customers will ask the vendor to change it.
What it means for US & EU software teams
For US teams, this is a product-requirements change with a deadline. Enterprise buyers with California staff will push compliance into procurement questionnaires within months: show us the human-review step, the notice template, the audit log, and the switch that turns off emotion analytics. Vendors that cannot answer will lose deals to those that can. California rarely stays alone — Illinois already regulates AI-analyzed video interviews — so a configurable, per-jurisdiction policy layer beats a one-off California fork.
For EU teams, the direction is familiar. The EU AI Act already prohibits emotion recognition in the workplace (with medical and safety exceptions) and treats AI used for promotion, termination and task allocation as high-risk, with human-oversight duties; GDPR Article 22 limits solely automated decisions with legal or similarly significant effects. A product built to EU standards is most of the way to California compliance. The reverse also holds: if you ship one HR product to both markets, design human oversight and emotion-free monitoring as the default, not as a regional patch.
The engineering lesson is that “human in the loop” must be provable. A rubber-stamp approve button will not show that a human independently corroborated the AI output. Teams need records of what the reviewer saw, which additional evidence they used, and when the worker was notified.
What to change before 2027
- Map every decision path. List features that score, rank, flag or recommend action on individual workers, and mark which can lead to discipline or termination.
- Add a corroboration step. Route those outputs into a review workflow where a named human records the additional evidence used. No automated status change without it.
- Generate worker notices. Build notice templates and delivery tracking for decisions where an automated system was used.
- Kill emotion inference by policy. Make sentiment, emotion and biometric-state analytics on employees switchable per jurisdiction, and off by default for California from January 1, 2027.
- Log for audit. Keep immutable records of model version, inputs, reviewer, decision and notice timestamps — the evidence a labor commissioner or PAGA plaintiff will request.
- Reuse EU AI Act work. Human-oversight and logging controls built for high-risk employment AI under the EU AI Act map directly onto SB 947.
Frequently asked questions
What did California sign on September 30, 2026?
Governor Gavin Newsom signed several workplace AI bills. SB 947, the No Robo Bosses Act, bars employers from relying solely on an automated decision system to discipline or terminate workers. AB 1883 bans AI-powered workplace surveillance that recognizes or predicts employees' emotional states or collects neural data. SB 951 updates Cal/WARN so employers disclose when mass layoffs are caused in substantial part by AI.
When do the California workplace AI laws take effect?
AB 1883 takes effect on January 1, 2027. SB 947 takes effect on July 1, 2027. Teams that build HR or workforce software have roughly three to nine months to ship the required controls before customers need them in production.
Does SB 947 ban AI in HR decisions altogether?
No. Employers can still use automated decision systems as an input. What SB 947 prohibits is relying solely on such a system for discipline or termination: a human must independently corroborate the output with other information, such as manager evaluations or personnel records, and affected workers must receive notice that an automated system was used.
What are the penalties under AB 1883?
AB 1883 carries civil penalties of up to $500 per violation, enforced by the state labor commissioner and public prosecutors, with potential exposure under California's Private Attorneys General Act (PAGA). Because penalties are per violation, a monitoring feature deployed across a large workforce can multiply exposure quickly.
Sources
Associated Press — California Gov. Gavin Newsom signs laws to protect workers from AI risks (September 30, 2026)
Bloomberg Government — California bans using AI to monitor workers’ brains, emotions (September 30, 2026)
CNBC — Newsom bans AI ‘robo bosses’ in landmark state law, reversing his earlier veto (September 30, 2026)
Ogletree Deakins — California Governor signs 3 bills targeting AI and workplace surveillance (October 2026)