California Just Made AI Layoffs a Matter of Public Record

On September 30, closing out a legislative session that sent nearly thirty AI-related bills to his desk, Governor Gavin Newsom signed SB 951, an amendment to California's WARN Act that does something no state or federal law has done before: it defines a mass layoff caused by artificial intelligence as its own legal category, with its own notice period, its own disclosure requirements, and its own public reporting channel.

The mechanism is narrow but pointed. Where California's existing WARN Act triggers at 50 layoffs within 30 days, SB 951 adds a lower, AI-specific bar: any relocation, mass layoff, or termination caused 'in whole or in substantial part' by an AI system or other automated technology, affecting 25 workers or 25 percent of a site's workforce — whichever number is smaller. Once that threshold is crossed, the standard 60-day notice period stretches to 90 days, and the notice itself must carry a header stating plainly that it covers a 'technology displacement,' name the automated system involved, and identify which job functions are being automated.

A paper trail where there wasn't one

The significance of SB 951 has little to do with the notice itself and everything to do with what happens after it's filed. California's Employment Development Department is now tasked with aggregating these AI-displacement notices into quarterly, statewide public summaries. That is a meaningfully different kind of data than what currently drives the national conversation about AI and jobs, which runs almost entirely on self-reported signals: earnings-call language, outplacement-firm surveys, and executives crediting or blaming AI for headcount decisions depending on which story serves the stock price that quarter. SB 951 produces something closer to a government-audited record, filed under penalty of the same back-pay liability that attaches to any other WARN Act violation, of when and where automation actually did the firing.

That distinction matters most to two audiences that have had almost nothing reliable to work with until now: plaintiffs' employment lawyers looking for a company's own description of why it cut staff, and investors trying to judge whether a company's AI-efficiency narrative on an earnings call matches what it is telling regulators about its workforce. A company that tells shareholders its margins are improving because of 'AI-driven productivity gains' while filing a Cal/WARN notice that attributes the same layoff to 'market conditions' has now created a discoverable inconsistency — one that didn't exist as a matter of public record before September 30.

This notice is for a technology displacement.

The loophole is built into the statute

The law's central vulnerability is also its central design choice: it never defines what counts as a layoff caused 'in substantial part' by AI. Employment lawyers tracking the bill have already flagged the phrase as an open invitation to both litigation and evasion. A company that wants to avoid the 90-day window, the public header, and the EDD's quarterly scorecard has an obvious option: describe the same cut as a restructuring, a cost-reduction initiative, or a response to demand softening, and let the automation do its work quietly. SB 951 gives California no mechanism to independently verify which label is accurate — it only requires that employers pick one and live with whichever internal documents, procurement records, or vendor contracts later surface if a worker or regulator contests the characterization.

That tension — a disclosure law whose central trigger is self-determined by the party with every incentive to avoid it — is not unique to SB 951, but it lands at a moment when the gap between companies' public AI narratives and their actual headcount decisions has become one of the most contested empirical questions in the labor market. A law that makes mislabeling a layoff legally risky, even without a bright-line test for what counts as AI-caused, shifts the burden in a way self-reporting never did: it puts a company's internal rationale for a layoff into a document that labor regulators, journalists, and opposing counsel can all request.

Why this bill, and not the more obvious one

SB 951 arrived alongside a more heavily covered sibling, SB 947, the 'No Robo Bosses Act,' which bars employers from letting an automated system fire or discipline a worker without human sign-off. SB 947 is itself a revived version of SB 7, a nearly identical bill Newsom vetoed in October of last year; this year's version survived by moving the human-notice requirement from before an AI-assisted decision to after it — a small procedural concession that was apparently enough to get a second-term governor with evident national ambitions to sign it. Both bills take effect in 2027, SB 951 on January 1 and SB 947 on July 1, giving California's enforcement apparatus and its EDD reporting pipeline more than a year to stand up before the first mandatory disclosures are due.

What makes SB 951 the more structurally important of the two is that it doesn't regulate how AI is used inside a company — it regulates what a company has to say, publicly, about the fact that it used AI at all. SB 947 governs conduct. SB 951 governs disclosure. And disclosure law, once a data trail exists, tends to outlive the political fights that created it: it gets cited in shareholder suits, pulled into congressional testimony, and used by researchers trying to settle arguments that had previously run entirely on anecdote. If a federal preemption push eventually narrows what states can do on AI safety and model-level rules, a labor-disclosure statute like SB 951 — built on a 1988-vintage WARN Act framework states have regulated for decades — sits on considerably steadier legal ground than the frontier-model transparency laws that have drawn the administration's sharpest fire.