California has decided an algorithm can’t fire you by itself. On September 30, 2026, Governor Gavin Newsom signed a package of first-in-the-nation bills governing how AI is used to discipline, monitor and lay off workers — headlined by SB 947, the No Robo Bosses Act, which blocks employers from relying solely on automated decision systems when terminating or disciplining staff. HR Dive’s compliance read calls it the most sweeping workplace AI measure enacted in the United States to date.
🔍 THE BOTTOM LINE: A decision made primarily by software now needs a human to corroborate it before it sticks. Employers must tell workers in writing when an automated system is being used on them, and mass layoffs “caused in substantial part” by AI trigger disclosure of the specific roles and system types involved. Business groups lobbied against the bill to the end — and their strongest argument, that the trigger phrase “primarily relies” is undefined, is exactly the ambiguity every jurisdiction regulating AI at work will now inherit.
What the four bills require
Per the governor’s office and HR Dive:
- SB 947 — No Robo Bosses Act: employers cannot rely solely on automated decision systems for discipline or termination. If a system “primarily” drives such a decision, a human must corroborate it using supporting evidence like evaluations and personnel files. Workers gain the right to request a description of the data the system used, and employers must give written notice of the system’s use. Effective July 1, 2027.
- SB 951: amends Cal/WARN so mass layoffs caused in substantial part by AI require written notice — including the occupations displaced and the category of AI system that caused it.
- AB 1883: bans workplace tools that infer workers’ emotional states or collect neural data — enforcement beginning January 1, 2027, extending a ban the EU already has in place.
- AB 1331: bars workplace surveillance tools in bathrooms.
The package follows the pattern this site has tracked as states converge on algorithmic accountability: regulate the decision, not the technology. Connecticut’s CART Act takes effect today with similar logic — automated employment decision-making no longer shields companies from discrimination claims, and employers have until October 2027 to build disclosure frameworks. Illinois already mandates AI-use disclosure in hiring but stops short of banning AI-only decisions — a gap SB 947 now fills first.
A veto reshaped into a narrower law
The signing is a second attempt. Newsom vetoed SB 7 in October 2025, calling the earlier bill unfocused. Author State Sen. Jerry McNerney brought it back with advance-notice provisions dropped and gig workers carved out, per CNBC’s account carried by Yahoo — concessions that did not win over California business groups, which pressed for another veto partly because the threshold phrase “primarily relies” went undefined. “No worker should ever be fired or disciplined by a robo boss,” McNerney said in a statement. “AI must remain a tool controlled by humans, not the other way around.”
Labor’s framing was celebratory: California Federation of Labor Unions president Lorena Gonzalez said Newsom “has shown the country how we can begin to put guardrails on AI at work to ensure human oversight over this technology, limit surveillance, and put workers first.” Employers, for their part, have until mid-2027 to figure out where their automated systems sit. Jackson Lewis’ analysis advises them to start by inventorying every tool that could qualify under the state’s deliberately broad ADS definition.
Our take: the interesting failure mode here isn’t enforcement — it’s the undefined phrase. “Primarily relies” is where the money and the lawyers will live: any employer that wants to keep automating can restructure paper trails so a nominally-human sign-off sits atop an automated decision, and proving that a human review was real rather than ceremonial will be hard. But watch what California actually did: it made the human the compliance artifact. In a labour market where tools increasingly act on people without their knowledge — Reuters reported in September on the Workday case, in which workers allege AI hiring screens filtered them out (paywalled in our check; named per policy, not counted as a loaded source) — forcing a named person to own each automated outcome is a quietly radical inversion: accountability by design. New Zealand has no equivalent on the horizon; the closest NZ institutions have come is guidance, not law. If the distributed, small-scale approach to AI governance works anywhere, it will be states like this one — and Connecticut, and Illinois — quietly building the rulebook the US Congress won’t.
📰 Sources
- Office of Governor Gavin Newsom — California’s nation-leading AI framework just got stronger (30 September 2026)
- HR Dive — California revamps AI protections for workers in flurry of bill signings
- Yahoo News — California No Robo Bosses Act bans AI-only worker firings
- Crypto Briefing — California signs first-in-the-nation laws putting guardrails on AI bosses
- CT Mirror — Connecticut AI and data privacy laws take effect in October
- California legislative information — SB 947, the No Robo Bosses Act — Jackson Lewis: California Passes No Robo Bosses Act (law-firm analysis; blocked to bots)
