Gavin Newsom just signed the law he killed a year ago. California employers can no longer let an algorithm fire someone without a human checking the work first.
Newsom vetoed this exact bill last October, calling it overly broad. On September 30, 2026, he signed it anyway. Senate Bill 947, the No Robo Bosses Act, bars California employers from relying primarily on an automated decision system to fire, discipline, or demote a worker. If a company wants to act on what the algorithm says, a human has to corroborate it first, using personnel files, manager evaluations, or peer review, not just rubber-stamp the output.
State Senator Jerry McNerney wrote the bill, reviving it this year after Newsom's veto of a nearly identical predecessor, SB 7, in October 2025. That earlier version cleared both chambers with overwhelming support and Newsom killed it anyway, writing that it "proposes overly broad restrictions" on how employers use these tools. Nothing about the policy changed much between the two versions. What changed was the politics: AI job displacement has become one of the defining anxieties of 2026, and Newsom has spent the year building a reputation as the country's most visible AI watchdog among governors.
The law also requires employers to give workers written notice when an automated system contributed to a termination or discipline decision, spelling out what data the system used, and to provide a human contact who can actually explain the decision. Workers get a private right of action. Violations carry civil penalties of $500 each, on top of possible punitive damages and attorney's fees, enforced by the state labor commissioner or by public prosecutors. None of it takes effect immediately. The law becomes operative July 1, 2027, giving employers and vendors about nine months to rebuild how their termination workflows actually work.
New York City got there first, but with a different tool. Local Law 144, in force since 2023, requires an independent bias audit before an employer can use an automated hiring tool, and the results have to be made public. It targets discrimination at the point of hiring. SB 947 targets something else entirely: who gets to make the call when someone is already employed and the question is whether to fire them. You can pass a bias audit and still let an algorithm cut someone loose with zero human review. California just made that illegal.
[Tech Layoffs Top 225,000 in 2026 as Companies Blame the Same AI They're Funding](https://startupfortune.com/tech-layoffs-top-225000-in-2026-as-companies-blame-the-same-ai-theyre-funding/)
Tech layoffs have topped 225,000 workers in 2026, with Challenger, Gray & Christmas data showing AI has become the leading stated reason for job cuts, ahead of overhiring or economic softness. Oracle's own numbers, a $2.8 billion restructuring bill alongside a $664 billion revenue backlog, show the pattern most starkly. - tech layoffs in 2026 blamed on AI - how many tech workers laid off this year
That distinction matters for anyone building HR software. Workday and HireVue-style vendors that sell scoring, flagging, or automated termination-recommendation tools to California employers now need a documented human-review step built into the product, not bolted on as a disclaimer. A tool that outputs a termination score is fine. A tool that triggers the termination is not, unless a person signs off using evidence beyond the algorithm's own output. That's a real product requirement, not a policy suggestion, and it carries a per-violation dollar figure attached to getting it wrong.
Newsom didn't sign SB 947 alone. According to reporting from the Associated Press, carried by outlets including the Washington Post, he also signed a law banning employers from using AI to infer a worker's emotional state from biometric data, a measure requiring written notice to workers when AI drives mass layoffs, and a separate law requiring operators of AI chatbots to run risk assessments before launch. He issued an executive order on top of all that, directing state agencies to keep calling the technology "artificial intelligence" rather than adopt the term "superintelligence," which the Trump administration had instructed U.S. diplomats to use instead.
Frankly, the chatbot risk-assessment piece and the biometric ban are going to get less attention than they deserve, because the robo-boss headline is the one that travels. But together they form the most complete state-level AI labor framework in the country, and they landed in the same week. For a founder building workforce AI tools in San Francisco, the lesson isn't just about SB 947's specific mechanics. It's that California intends to keep legislating here every single session, and a veto one year doesn't mean the fight is over. It means it's coming back next year, usually larger.
Other states are watching the same thing. Illinois and Colorado have already built different versions of anti-discrimination liability around automated employment tools, and California's own law stacks on top of existing FEHA and CCPA obligations rather than replacing them. Compliance with one doesn't satisfy the other. If SB 947 becomes the template other legislatures borrow from, as NYC's bias-audit model already has, the HR-tech compliance map in 2027 looks nothing like it does today.
Also read: Bank of England Governor Warns AI Valuations Could Face a Sharper Correction • Gemini 4 Argon Posts the Lowest Hallucination Rate of Any Top AI Model • Huawei warns of more smartphone price hikes as the memory chip crunch bites
This article is posted in AI News, check it out for more related stories.
California built a real-time dashboard to track whether AI is actually killing jobs California launched CAIT, the nation's first statewide AI-unemployment tracker, on June 25, 2026, linking real-time unemployment insurance data to AI-exposure measures by occupation. The initial findings show no aggregate surge, but college-educated workers in high-AI-exposure roles have seen elevated claims since ChatGPT-3.5 launched in 2022... - AI job loss tracking dashboard California - real time employment impact of artificial
Join the discussion #
Open in the community → Almost there. Sign in and your reply posts straight away.