California bans emotion-reading AI at work and requires a human before you're fired by algorithm

By Zendoric · October 2, 2026.

California Governor Gavin Newsom signed on September 30 a package of three laws limiting the use of artificial intelligence to monitor and manage employees. AB 1883 bans workplace surveillance tools that use AI to infer a worker's emotional state or to collect 'neural data' —information about their central or peripheral nervous system, for example through wearable devices—. SB 947, dubbed the 'No Robo Bosses Act', prohibits firing or disciplining an employee based solely on an automated decision system (ADS): from July 1, 2027 there will always have to be a human supervisor who independently corroborates that decision. AB 1331, for its part, bars the use of surveillance tools inside workplace restrooms, with limited exceptions.

Violations of AB 1883 and AB 1331 can cost up to $500 per case, according to the employment law firm Ogletree Deakins in its reading of the legal text. The laws do not create a direct private right of action, but they can be enforced through PAGA (the Private Attorneys General Act), the California mechanism that allows any worker to sue on the state's behalf over Labor Code violations. SB 947 also adds two little-discussed safeguards: it prohibits using an ADS to infer an employee's protected characteristics (national origin, disability, etc.) and to retaliate against someone exercising a labor right, and it requires after-the-fact notice when an employment decision relied on an automated system.

The model is not new: the European Union's AI Act already bans AI emotion recognition in the workplace and in education, and the firm's analysis places the California rule in that same lineage. Nor is it the first attempt: SB 947 is a revised version of a bill Newsom himself vetoed in 2025, a sign that the clash between unions and the tech industry has not been resolved, only reframed until a text emerged that the governor would sign. The final version is in fact narrower than the early drafts: it explicitly allows the use of surveillance for safety reasons —such as sensors that detect driver fatigue— and reaches only AI-generated predictions, not conventional video surveillance or time tracking.

The measure does not arrive alone. Around the same dates Newsom signed two executive orders: N-10-26, of September 30, which requires the state administration to keep calling AI 'artificial intelligence' regardless of how the federal government rebrands it, and N-9-26, of September 18, which explores greater oversight of frontier model development, including a possible mandatory 'kill switch'. Together they sketch a state that, in the absence of federal legislation on AI in the workplace —and with none expected in the short term, according to the article itself—, is building its own regulatory framework piece by piece. Colorado, Connecticut, Illinois, New York (city) and Texas already have similar laws on automated decisions in hiring or firing; the risk is a patchwork of fifty different rules that ends up pushing, sooner or later, toward a federal standard.

Our reading: this is exactly the kind of short-term friction that should not be downplayed. Algorithmic surveillance —reading an employee's pulse, tone of voice or tiredness to decide whether to promote or fire them— is one of the fronts where AI concentrates power most quietly, because it happens inside the company, without the public scrutiny that frontier models do receive. That the law requires a human to review every automated dismissal, rather than a blanket ban on the technology, is a reasonable design: it targets the specific abuse (the decision with no one accountable) without prohibiting AI applications that do deliver real value, such as fatigue detection that can save lives on the road. It is governance aimed at verified harm, not generic panic, and it is the kind of guardrail we need to get past during the hard transition before the abundance promised by AI reaches workplaces in the form of fewer routine tasks and more human judgment, not fewer rights.