California Governor Gavin Newsom signed a landmark AI law that prohibits Golden State employers from relying solely on artificial intelligence to fire or discipline workers.
SB 947, known as the “No Robo Bosses Act,” prohibits businesses in the state from exclusively using “automated decision-making systems” to discipline or fire employees. It also limits the ability to use AI as a primary tool in such decision-making.
Under the bill, employers that rely “primarily” on AI output to make firing or disciplinary decisions would be required to have human reviewers back up their decisions with additional information such as manager reviews, peer reviews, and personnel files. In addition, affected employees must be provided with written notice that AI was “primarily used” in termination or disciplinary decisions, an explanation of the employee data used in the system, and a human contact point who can explain the decision in more detail.
California Sen. Jerry McNerney, the bill’s author, told CNBC: “No worker should ever be fired or punished by a machine, AI or not. Artificial intelligence systems have the potential to increase productivity, but they also make mistakes, make bad decisions, and show bias.” “AI should remain a tool controlled by humans, not the other way around.”
McNerney, a Democrat, first introduced the legislation in 2025 following a concerted push by organized labor in the state to install guardrails that would prevent management from using AI to take action against workers.
“When working people come together, we get results. Workers across California have been demanding that the state take the lead in regulating AI in the workplace, and today we are seeing that begin to happen,” Lorena Gonzalez, president of the California Federation of Labor Unions, AFL-CIO, and lead sponsor of the Anti-Roboboss Act, said in a statement after Newsom signed the law. “Today, California workers and unions changed the national narrative on how Americans can fight back and win against the takeover of our jobs and workplaces by AI.”
Newsom has taken a flurry of recent actions on AI, including signing a wide-ranging executive order earlier this month to address the potential existential risks posed by AI models “before it’s too late” and steps to provide a national framework for independent evaluation and auditing of AI models. Newsom said the broader EO is an abdication of the federal government’s “responsibility to protect the American people.”
The United States leads the world in the adoption of automated management software and systems, according to an OECD study released late last year. The study found that algorithmic management software is widely adopted in many countries, but nowhere more widely than in the United States, with 90% of managers saying their companies have at least one tool in place to “coach, monitor, and evaluate employees.”
Meta is facing a lawsuit filed in July. Former employees allege that the AI-assisted system was used to rank and select employees for layoffs, unfairly influencing employees who took sick or family leave. Mr. Mehta denies the allegations.
A May survey released by United for Respect, a nonprofit organization with roots in the labor movement and the retail industry in particular, found that employees at Walmart and Amazon are increasingly concerned that personnel decisions are being automated. The nonprofit group unsuccessfully tried last summer to persuade Walmart shareholders to approve a measure that would require company management to provide more disclosure about its use of AI to employees.
California’s anti-roboboss law hit a wall when Newsom vetoed it in October, even though it passed with overwhelming support in both state legislatures. Among his concerns, Newsom cited requirements that would require companies to give advance warning to workers if AI systems are being used that could affect working conditions.
In his briefing, the governor said he “shares the authors’ concerns that, in some cases, unregulated use of ADS (automated decision-making software) by employers could be harmful to workers,” and foreshadowed his future willingness to accept similar legislation. “But rather than addressing the specific ways that employers misuse this technology, this bill imposes unfocused notification requirements on any company that uses the most innocuous of tools,” he continued.
When McNerney reintroduced the bill in February after resistance from Newsom, he removed the advance notice requirement and removed language extending the law’s protections to gig workers. This drew heavy criticism from ride-hailing giants like Uber and Lyft. Although these amendments have somewhat slowed the intense lobbying by tech and trade groups that helped kill the 2025 law, California business leaders still rallied against SB 947 in the weeks leading up to its enactment.
“The bill’s obligations generally apply when an employer relies ‘primarily on’ an automated decision-making system, but that key term is never defined. Employers will be asked when technology moves from solely informing decisions to being the primary basis for decisions. “We are not given an objective standard by which to judge the bill,” Robert Singleton, senior director of policy and communications for California and the Western United States at the Progressive Chamber, wrote in a letter to Newsom earlier this month, urging the governor to veto the bill. “Uncertainty about whether regular tools qualify as regulated automated decision-making systems may deter employers from using technology that improves consistency, identifies safety risks, or enables managers to make better-informed decisions,” he added.
First law in Japan to allow citizens to turn on AI before elections
While the aforementioned changes were made specifically to win Newsom’s approval, NoRobobos supporters also benefited from the fact that the bill reached the governor’s desk at a time when public distrust over the introduction and continued presence of AI in the workplace is at an all-time high. A July Gallup poll found that 39% of Americans believe AI will do more harm than good, an 8% increase from the 31% who felt the same way in 2025. A Pew Research survey released in August similarly found that 71% of Americans believe AI will “take away people’s jobs,” an increase of about 7% from about two years ago.
This trend is growing as Newsom considers a presidential run in 2028, where AI policy is expected to be a hot issue, especially among Democratic candidates. But Democrats are not alone on this issue. Earlier this month, California Republican gubernatorial candidate Steve Hilton announced his full support for legislation banning roboboss.
“The California Chamber of Commerce…says that rules requiring human review and basic accountability are too burdensome for employers. They’re ridiculous for saying that,” Hilton wrote in an open letter to Gonzalez. “In some ways, I don’t think SB 947 goes far enough,” he wrote. “Employers should not be allowed to use AI to determine whether someone will be fired, demoted, lose regular hours, or be locked out of programs they rely on for income.”
California businesses will be the first to navigate a unique regulatory framework that establishes an almost unprecedented level of oversight over the use of AI in the workplace. Throughout the past five years, some local regulations have imposed limits on how AI can be used in employment decisions, perhaps most notably Illinois Public Law 103-0804, which also requires employers to notify workers when using AI for certain employment purposes. However, these limitations do not go as far as current California law. Even the Illinois law that went into effect in January of this year stops short of completely banning the implementation of proxy digital systems for employment decisions.
Although the Roboboss Act is the first law of its kind to be passed, many other bills aimed at serving exactly the same purpose currently remain pending across the United States. In June, Sens. Ed Markey (D-Mass.) and Brian Schatz (D-Hawaii), both Democrats, introduced a federal bill of the same name that would prohibit employers from “relying on automated decision-making systems to make work-related decisions.” Markey’s bill was effectively dead on arrival in a Congress slow to develop any legislation regulating AI. A number of states, including New York, Louisiana, and New Jersey, have proposed similar bills but have not yet taken voluntary action.
Earlier this year, New York became the first state in the nation to enact a law banning the construction of new AI data centers.
California’s action may just be the boost other states need to tackle AI in the workplace legislation.
The left-leaning freedom nonprofit Electronic Frontier Foundation said in a statement to CNBC that signing the law is “a strong step toward providing necessary protections for workers in workplaces that use automated decision-making systems” and that it plans to continue working with labor organizations to “advance policies that protect human dignity in the workplace.”
