California's 'No Robo Bosses Act' bans fully automated firings

California has outlawed firing workers entirely by algorithm. SB 947, signed into law under the informal name "No Robo Bosses Act," prohibits employers from outsourcing disciplinary and termination decisions to what the bill calls automated decision systems (ADS), and it takes effect on July 1, 2027.

The law does not ban AI from the process. It bans the absence of a human. Employers must check any AI decision on discipline or termination and give the employee a description of the reasons behind the decision. That description has to cover the data the system used — personnel files, evaluations, work product, peer reviews, and "witness interviews."

The bill draws a hard line when the human check fails. "If an employer cannot corroborate the ADS output or the human reviewer has concluded that the ADS output is inaccurate, incomplete, or misleading, the employer shall not use the ADS output to make a disciplinary or termination decision," the bill states. In plain terms: if the human can't back up the machine, the machine's verdict is dead on arrival.

The stakes are not hypothetical. AI already sits inside hiring, performance review, and layoff decisions at a scale most workers never see.

How much of HR is already automated?

An April survey by MyPerfectResume of 1,000 HR professionals puts numbers on it. 73% said they use AI to make hiring decisions. 52% said they use AI to inform "restructuring and role planning" — the corporate language for deciding which jobs survive.

The screening goes deeper than resume matching. Over half of respondents, 51%, use AI to "flag 'risky' candidates." On the exit side, only 26% said they don't use AI to make decisions about layoffs. That means nearly three-quarters of surveyed HR professionals leave at least some layoff decisions, or parts of them, to software.

The takeaway: AI is already making consequential calls about people's livelihoods, and most employees have no visibility into it.

What does the new law actually change?

SB 947 creates two concrete obligations for California employers. First, a human must review any ADS output before it drives discipline or termination. Second, the employee gets an explanation — the reasons for the decision and the data behind it.

The law also adds a transparency requirement for larger workforce actions. According to a brief on the governor's website, employers must clarify whether "a mass layoff, relocation, or termination is caused by an AI system."

Enforcement runs through the California Labor Commissioner. Affected employees can file complaints if they believe the process wasn't followed correctly in a firing decision.

Note what the law does not do. It doesn't fully prohibit automated systems for these determinations. An employer can still lean on AI heavily, as long as a human signs off and the worker gets an explanation. The law regulates the checkpoint, not the pipeline.

Why does one state law matter nationally?

Federal AI regulation remains thin on worker protections — the ZDNET report characterizes it as "less specific, mired in voluntary agreements with labs." Into that gap, states are writing the first real rules.

The approaches diverge sharply. Colorado and Connecticut have opted for mandating disclosure of AI use in hiring and workplace processes rather than actual intervention. California and New York have historically pushed stricter laws. And this summer, Illinois passed a landmark bill becoming the first state to require third-party audits of frontier labs.

Lily Li, a lawyer specializing in AI and founder of Metaverse Law, told ZDNET she sees California's law setting the precedent for other states. The state's size gives it outsized influence: companies that build one compliance process for California tend to apply it everywhere.

Will the human review requirement hold up in practice?

Li anticipates a specific failure mode: companies gaming the review process. "We could see employers engaging in check-the-box reviews rather than meaningful human review, to avoid the operational costs of an additional layer of review," she said.

Real review costs money. A human who genuinely re-examines the data behind every AI-driven discipline decision adds a layer of labor that some employers will want to minimize. A signature on a form is cheaper than an investigation.

Li sees one counterweight built into the law. "The ability for an employee to request an explanation is a helpful check against this, however," she said. If workers can demand the reasons and the underlying data, a hollow review becomes easier to expose.

The likely consequence, as Li's analysis implies: the burden of policing unvetted AI-driven firings may fall on complainants contesting their own terminations in court. The law creates the right to challenge; it doesn't guarantee the challenge is easy.

What happens between now and 2027?

The law doesn't take effect until July 1, 2027, leaving a long runway. Employers have time to build review processes — or, if Li's prediction holds, to build the thinnest review processes that survive legal scrutiny.

The structural problem remains. Protections like California's will stay patchwork across the country until some version of them is adopted at the federal level. A worker in Sacramento gets an explanation right and a human-review guarantee. A worker in a disclosure-only state gets, at most, a notice that AI was involved.

The bottom line: California has defined the current ceiling for worker protection against algorithmic firing — and the real test will be whether human review means human judgment or a rubber stamp.