California bans AI-only firing decisions and bathroom surveillance

California employers can no longer rely solely on artificial intelligence to fire workers, under new legislation Governor Gavin Newsom signed into law this week. The measures, announced by Newsom, also block AI-powered surveillance of workplace bathrooms and bar attorneys from "fully handing over core legal work" such as drafting briefs to AI systems.

The package arrives as some of the first legislation in the nation that gives employees enforceable rights over how AI can be deployed in their workplaces. That matters far beyond California: the state routinely sets de facto national standards for technology policy, and the federal government has signaled it will not fill the gap. Newsom made that point in blunt terms, calling recent statements from the Trump administration that the AI industry can self-regulate "bulls*it," and adding: "we have to do a lot more in the absence of federal leadership."

"AI should expand opportunity — not come at the expense of workers and families," Newsom said in a statement. "As this technology reshapes the workplace, California is putting people at the center."

What the new laws actually do

Divided into multiple bills, the legislation targets two distinct risks: automated discipline and automated monitoring. First, it prohibits employers from relying only on AI when making a disciplinary action or termination decision. A human decision-maker, with actual authority over the outcome, must remain in the loop when a person's job is on the line.

Second, it forces transparency when AI drives workforce reductions. Companies must disclose if any mass layoff, relocation or termination is caused by an AI system. That requirement gives employees and regulators a paper trail where none previously existed, and it creates a concrete legal exposure for firms that let opaque models drive staffing cuts.

Third, the legislation protects workers against surveillance tools installed in workplace bathrooms — a measure that speaks to how far some employers have pushed monitoring technology marketed as productivity tooling.

Fourth, it draws a line inside the legal profession itself. The law prohibits lawyers from "fully handing over core legal work," including drafting briefs, to AI. The provision reflects growing concern that generative models, which remain prone to fabrication, could compromise the quality of filings that courts and clients depend on.

Notably, the legislation does not stop companies from using AI to replace workers outright. It is aimed at how companies deploy the technology in job cuts and monitoring, not at automation-driven displacement itself. That leaves the more politically explosive question — what happens to jobs that AI simply eliminates — outside the reach of these bills.

Why now: Meta complaints and manager behavior

The legislative push did not emerge in a vacuum. California's new laws may have been motivated in part by complaints from Meta employees that the company used biased AI tools that "disproportionately selected" people who took medical leave as candidates to be laid off. If accurate, that pattern would sit at the intersection of algorithmic bias and protected leave — precisely the scenario the new termination provisions are designed to surface and constrain.

The practice is not rare. One in four managers use AI to help decide which employees to cut "often or all the time," according to human resources site HR Dive. That figure gives the legislation a concrete target: automated input into layoff decisions has moved from experiment to routine practice, while legal accountability for those decisions has lagged.

For employees, the stakes are straightforward. An AI system trained on historical performance data can inherit and amplify bias against workers who took parental leave, medical leave, or who belong to underrepresented groups. Before this law, a worker fired on the recommendation of such a system had little visibility into the mechanism behind the decision, and little statutory footing to challenge it.

A terminology fight with Washington

Newsom, widely discussed as a potential 2028 presidential candidate, paired the worker protections with an executive order requiring state agencies to continue referring to the technology as "artificial intelligence" rather than "super intelligence." The order directly counters a recent mandate from the Trump administration, which directed some federal employees to adopt the newer term.

The naming dispute is small on its face but signals a larger divergence in how Sacramento and Washington frame the technology. The federal push toward "super intelligence" language emphasizes ambition and capability. Newsom's order keeps the framing sober and technical. He is positioning California as a leader in AI regulation, likening the technology to the airline industry in terms of its potential dangers to the public — an analogy that implies licensing, inspection regimes and liability rules rather than voluntary industry commitments.

The regulatory vacuum California is filling

The context for these bills is a federal government that has moved toward deregulation. The Trump administration's stated position that the AI industry can police itself, which Newsom dismissed outright, leaves states as the primary venue for binding AI rules affecting workers. California's legislation becomes a test case for what enforceable workplace AI rights look like in practice: not bans on the technology, but requirements for human judgment, disclosure and limits on invasive monitoring.

The design choice is telling. Rather than regulating AI capabilities in the abstract, the bills attach obligations to specific consequential moments — firings, mass layoffs, disciplinary actions, legal filings, bathroom monitoring. That approach maps to how workers actually experience the technology: not as a distant research frontier but as a system that scores them, flags them and, in some cases, selects them for termination.

Compliance questions will follow. Companies operating in California will need to document when AI systems contribute to termination decisions, establish review processes that satisfy the human-oversight requirement, and audit any monitoring tools that touch physical spaces. Firms that already use algorithmic ranking in workforce reductions — the Meta complaints suggest at least one major employer did — face the most direct adjustment.

What comes next

The transparency requirement around AI-driven layoffs may prove the most consequential provision, because it creates data where none existed. Once companies must report when a mass layoff or termination is caused by an AI system, regulators, journalists and plaintiffs' attorneys gain visibility into how often algorithms drive job cuts — and whether the patterns Meta employees alleged recur elsewhere.

Newsom's airline-industry framing points to where he intends to go: toward a broader inspection-and-liability regime for AI, built state by state if the federal government declines to act. With one in four managers already consulting AI on layoff decisions, the question for the coming years is whether other states follow California's template or leave workers in jurisdictions with no equivalent protections.