California Gov. Gavin Newsom signed SB 947 on Wednesday, requiring employers to obtain human verification before using artificial intelligence as the primary basis for firing or disciplining workers. It’s the first statewide restriction of its kind to reach the books, though it came only after Newsom rejected an earlier version and lawmakers removed provisions that had drawn fierce opposition from the rideshare industry.
When employers rely primarily on AI output to terminate or discipline workers, the law requires a human reviewer to corroborate the decision with additional information, such as managerial evaluations, peer reviews, or personnel files. Affected employees must receive written notice that AI played the primary role, a description of the employee data the system analyzed, and contact information for someone who can explain the reasoning.
California State Senator Jerry McNerney, the bill’s author, described the measure as a guardrail against machine error. “No worker should ever be fired or disciplined by a machine, AI or not,” McNerney told CNBC. “Artificial intelligence systems have the potential to boost productivity, but they’ve also made errors and misjudgments and exhibited bias. AI must remain a tool controlled by humans, not the other way around.”
The Business Objection
Employers and trade groups fought the bill vigorously. Robert Singleton, senior director of policy and public affairs for the Chamber of Progress’ California and US West operations, warned Newsom in a pre-signature letter that lawmakers had left the law’s central term—“primarily relies”—undefined. He argued that the ambiguity would discourage firms from deploying technology that improves consistency, identifies safety risks, or helps managers make better-informed decisions. Without an objective standard, businesses would face legal uncertainty simply for using tools intended to improve decision quality.
Newsom initially shared some of those concerns. When he vetoed an earlier version in October, he criticized the bill for imposing “unfocused notification requirements on any business using even the most innocuous tools.” He objected specifically to an advance-notice requirement that would have forced companies to alert workers whenever an AI system could affect work conditions. The governor acknowledged that “in certain cases unregulated use of ADS [automated decision-making software] by employers can be harmful to workers,” but argued that the first draft addressed the problem too broadly.
McNerney scaled back the proposal in response. In February, he removed the advance-notice requirement entirely and cut language extending protections to gig workers—a critical concession meant to defuse opposition from Uber and Lyft. Business groups still opposed the bill.
A Fractured Political Landscape
The bill’s route to Newsom’s desk exposed unusual political divisions. Steve Hilton, California’s Republican nominee for governor, broke with the state Chamber of Commerce and endorsed the measure. “The California Chamber…says rules requiring human review and basic accountability are too burdensome for employers. What a ridiculous thing for them to say,” Hilton wrote in an open letter to labor leader Lorena Gonzalez. “In some ways, I don’t think SB 947 goes far enough,” he added, arguing employers shouldn’t be allowed to use AI to decide whether someone is fired, demoted, loses regular hours, or is shut out of income-generating programs.
Gonzalez, president of the California Federation of Labor Unions, AFL-CIO and lead sponsor of the act, celebrated the signature as validation of labor’s organizing power. “When working people organize, we get results,” she said. “Workers across California have demanded that our state lead the way in regulating AI in our workplaces.”
National Context and Precedent
CNBC identified SB 947 as the first law of its kind to take effect at the state level. Illinois passed Public Act 103-0804, which requires notice when employers use AI for specified employment purposes, but that law—which took effect in January—doesn’t bar automated systems from making employment decisions outright. Similar measures have failed in New York, Louisiana, and New Jersey. New York became the first state to ban construction of new AI data centers earlier in the year.
At the federal level, U.S. Senators Ed Markey (D.-Mass.) and Brian Schatz (D.-Hawaii) introduced legislation in June with the same name, designed to bar employers from “relying on automated decision systems to make work-related decisions.” CNBC described that effort as “effectively dead on arrival.”
The Data Behind the Push
An OECD survey published late the previous year found that 90% of U.S. managers reported their firms had adopted at least one tool to “instruct, monitor or evaluate workers”—a higher adoption rate than anywhere else globally. Meta faced a July lawsuit from former employees, who alleged that AI-assisted systems ranked and selected workers for layoffs in ways that disproportionately affected those who’d taken medical or family leave. Meta denied the allegations.
Walmart and Amazon workers grew increasingly concerned about automated HR decisions, according to a May survey by United for Respect, a nonprofit with roots in the labor movement. The organization failed the previous summer to persuade Walmart shareholders to demand greater disclosure about AI use in workforce management.
Public sentiment shifted, too. A July Gallup poll found 39% of Americans believed AI did more harm than good, up from 31% in 2025. A Pew Research study released in August found 71% thought AI would “take people’s jobs,” an increase of roughly 7% from around two years earlier.
Why This Matters:
SB 947 reflects growing tension between technological adoption and regulatory constraint. The law imposes requirements justified by concerns about algorithmic bias and opacity, while employers now face legal and operational complexity when deploying tools designed to improve hiring and personnel consistency. The undefined term “primarily relies” leaves businesses guessing about compliance, potentially chilling investment in automation that could enhance accuracy and reduce favoritism. The measure also reflects a broader pattern: state-level regulation of AI employment decisions, driven partly by labor mobilization and partly by political figures weighing AI as a 2028 campaign issue.
Whether the human-review requirement meaningfully protects workers or simply adds procedural steps remains untested. What’s clear is that California—and potentially other states—has chosen regulation over market competition as the way to address algorithmic accountability, shifting the burden to employers to prove their systems aren’t the principal deciders in personnel matters.