AI Replacing Workers: Newsom's Order and Its Impact

California Governor Gavin Newsom signed an executive order on May 21, 2026, directing state agencies to assess AI's impact on employment and propose ways to blunt job losses, a direct response to labor unions and advocacy groups pushing for protective measures as AI spreads through everyday business operations.

What Happened

The order requires a study to identify which jobs are most at risk from AI and to develop retraining and support strategies for affected workers. It doesn't ban or restrict AI deployment. It requires the state to map the risk first.

That distinction matters: California is one of the largest state economies to formally direct government resources at measuring AI's job impact, rather than reacting only after layoffs happen.

How This Affects Everyday People

For workers in roles built around repetitive tasks or data processing (a factory worker in Los Angeles, for instance, watching AI-powered machines take over line work), the order signals the state is at least tracking the risk, even if it can't yet promise protection. Families are left weighing the same uncertainty: whether the careers they've counted on will still exist in a decade.

Communities concentrated in automation-exposed industries face the added risk of reduced economic mobility if the retraining piece of Newsom's order doesn't move as fast as the job losses do.

The Bigger Picture

Newsom's order follows the EU's AI Act, a comprehensive regulatory framework introduced earlier this year, and lines up with the U.S. Department of Labor's own tracking of AI-related layoffs via resources like our AI layoffs tracker. Government bodies are increasingly treating AI's labor impact as something to measure and manage, not just observe.

What You Can Do

• Follow how AI is affecting your industry with our explainers on AI and jobs.
• Reskill toward roles that lean on human empathy or creativity: see our guide to jobs AI can't replace.
• Use community and local-government retraining programs as they roll out under orders like this one.
• Weigh in with your local representatives on whether California's approach should extend to your state.

The Bottom Line

Newsom's order is a data-gathering step, not a guarantee. It commits California to actually measuring which jobs AI puts at risk, which is more than most states have done, but the real test is whether the retraining programs it calls for arrive before the layoffs do.

What California's Own Numbers Show

Newsom's order asks state agencies to map AI's job impact. Our own AI layoffs tracker already has a partial answer for California specifically. Of the 38 companies we've verified citing AI in layoffs since 2023, 11 are California-based, including Meta (8,000 roles, April 2026), Cisco (4,000 roles, May 2026), Salesforce (4,000 roles, September 2025), Block/Square (4,000 roles, February 2026), and Intuit (3,000 roles, May 2026). Together those 11 California companies account for at least 28,200 of the announced roles in our dataset, a floor, since several entries didn't disclose an exact headcount.

That concentration is exactly why California is a reasonable place to start measuring AI's labor impact before other states: it's home to a disproportionate share of the companies making these announcements in the first place. Whether Newsom's order translates into retraining funding that keeps pace with that number is the open question our tracker will keep documenting as new entries come in.

Salesforce's cut is a useful case study for what the order is trying to get ahead of. CEO Marc Benioff confirmed 4,000 customer-support roles were eliminated after the company's Agentforce AI agents began handling about half of all customer interactions, shrinking support headcount from 9,000 to 5,000, with some staff redeployed into sales rather than let go outright. That kind of internal redeployment, not just outright termination, is one of the retraining outcomes Newsom's order is explicitly trying to encourage statewide.

Not every California company in our tracker has offered that kind of internal path. Pinterest's 15% workforce reduction in January 2026 was framed as redirecting resources toward AI-focused roles and teams, but the company didn't disclose how many of the affected 700 workers were offered a path into those new roles versus simply let go. That gap between companies that redeploy and companies that don't disclose is precisely the kind of detail Newsom's order would require state agencies to start tracking.

The Order Sets Four Concrete Deadlines

Executive Order N-6-26 does more than call for a study. It sets four specific deadlines, and each one falls to a different state agency.

• Within 90 days, the Employment Development Department (EDD) must launch a public dashboard tracking AI's effect on employment by sector, built from unemployment insurance data the state already collects.
• The EDD then reports twice a year through the end of 2027 on what employers themselves say about AI's role in hiring and workforce decisions.
• Within 180 days, the Labor and Workforce Development Agency must recommend updates to California's Worker Adjustment and Retraining Notification (WARN) Act, the law that requires 60 days' notice before a mass layoff, so it flags AI-driven cuts sooner.
• By October 15, 2026, the same agency has to report on how union contracts are handling AI, and how workers currently get a say before an employer adopts it.

None of the four items changes what an employer can legally do to a workforce today. The order states plainly that it creates no new enforceable right a worker could take to court, the same measure-first limitation raised above, now with actual dates attached.

Why Estimating AI's Job Risk Is Harder than It Sounds

Economists have tried to size automation's threat to jobs before, and the track record is mixed. A widely cited 2013 Oxford study by Carl Benedikt Frey and Michael Osborne put 47% of U.S. jobs at high risk of automation within a decade or two. That figure hasn't held up well. Most of the occupations the study flagged as easiest to automate still employ roughly as many people as they did in 2013, and later researchers found the model scored entire job titles as if every task inside them were equally exposed to a machine, according to a 2022 review by the Information Technology and Innovation Foundation.

The EDD's new dashboard is built to sidestep that specific mistake. Instead of one projection made years in advance, it tracks actual unemployment insurance claims and employer-reported hiring changes as they accumulate, a running count rather than a forecast. Whether that produces a clearer picture than the Frey and Osborne model won't be testable until the dashboard has a few quarters of data behind it.

The Real Test Behind the "First-of-Its-Kind" Label

California's own announcement calls the order "first-of-its-kind." That claim rests on a specific distinction: a standing, government-run dashboard that keeps updating, not a one-time report or a task force with an expiration date.

It also raises the bar for the EDD's own 90-day deadline. A dashboard that launches late, or launches without real unemployment data behind it, would cost California the exact distinction its announcement claims. Whether that deadline holds is now a fact regulators, unions, and affected workers can check against a public record, rather than take on faith.

Further Reading

This story is based on original reporting via Google News: "Newsom warns AI replacing workers, signs order."