The name does most of the work. California's new "No Robo Bosses Act," signed by Governor Gavin Newsom on 30 September, sounds like a slogan, and it is one. But it is also the first law in the country to say that a machine cannot be the one to end your job. Behind the headline sit three more statutes, a cluster of compliance headaches, and a single undefined word that will quietly decide whether any of it bites.
Start with why it exists. A survey this year of 1,000 HR professionals found that 73 percent already use AI to help make hiring decisions and 52 percent use it to plan "restructuring." Only 26 percent said they do not use AI at all when deciding on layoffs. Into that reality steps SB 947, which bars employers from relying "solely" on an automated decision system to discipline or fire someone. If a tool is the primary basis for the call, a human must corroborate it against real evidence, the worker must be told afterward, and they can demand a description of the data used against them.
It does not stand alone. Three companion laws take effect on 1 January 2027, with the No Robo Bosses Act following on 1 July. AB 1331 bans workplace surveillance in bathrooms. AB 1883 forbids employers from using AI to infer an employee's emotions or to harvest "neural data." And SB 951 amends California's layoff-notice law so that a mass redundancy caused by automation must now carry the words "This notice is for a technology displacement," naming the specific system that replaced the workers. California becomes only the second state, after Connecticut, to force that kind of admission into the open.
The tensions are in the drafting. The No Robo Bosses Act leans on the idea of an employer relying "primarily" on a system, a word it uses six times and never defines. Lawyers expect that gap to be filled by cautious guessing until regulators or judges settle it. Lily Li, an AI lawyer who founded Metaverse Law, warns of the obvious workaround: "We could see employers engaging in check-the-box reviews rather than meaningful human review." A manager who reads a dashboard score and signs off has corroborated nothing, which is precisely why the statute shifts the burden of proof onto the employer once a worker shows a system was involved.
This is Newsom's second attempt. He vetoed a broader version, SB 7, in 2025, objecting that its advance-notice rules swept too wide. The 2026 law drops those requirements and excludes gig workers, which is narrower but passable. It carries civil penalties of 500 dollars per violation, enforced by the Labor Commissioner, with exposure that scales by the number of affected employees.
The larger significance is the precedent. Federal AI policy remains a patchwork of voluntary pledges, so the real rulemaking is happening state by state: Illinois now requires third-party audits of frontier labs, New York has added AI to the reasons an employer can give for a layoff, and California tends to set the template others copy. For multistate employers, the safest bet is to apply the human-review rule everywhere rather than carve out California. The open question is whether a mandated human in the loop becomes a genuine check or just a signature at the bottom of the machine's decision.