SB 951 Cleared Committee 5-0. The EDD Will Name Your Next Hire 90 Days Early.
California's SB 951 would force employers to publish AI-displacement notices 90 days early, naming workers, functions, and AI vendors. A sourcing goldmine.
On June 10, 2026, California's SB 951, the Worker Technological Displacement Act, passed out of committee 5-0 and was re-referred to the Committee on P. & C.P. If it survives the next round, the EDD will start publishing something no sourcing tool has ever had: a structured, government-mandated dataset of named workers, named job functions, and named AI vendors, 90 days before separation. That is not a layoff tracker. That is a sourcing feed.
Most recruiters reading this still treat WARN as a reactive signal. Layoff hits the news, scroll LinkedIn for "Open to Work" badges, send 200 templated messages, hope. SB 951 inverts that loop. The signal arrives before the badge, before the goodbye post, before the hiring manager at the next employer has even written the JD.
Here is what is actually in the bill, and how to build a pipeline against it now.
What SB 951 actually requires
The bill, authored by Sen. Sabrina Cervantes Reyes and backed publicly by the California Labor Federation as the "AI Job Killer Notice" bill, sets a tougher threshold than Cal-WARN. Standard Cal-WARN kicks in at 60 days and 50 workers. SB 951 kicks in at 90 days and 25 workers, or 25% of the workforce, whichever is less. That last clause matters for small AI-native teams cutting human support layers.
The required notice contents read like a recruiter's wishlist:
- Name and address of the employment site
- Contact official at the employer
- Expected separation dates
- Number and classification of workers being replaced
- The specific job functions being automated
- The AI system used
- The vendor that developed or sold it
There is also a parallel "technology hiring disruption notice" for when an employer permanently stops hiring for a role because AI absorbed the work. That filing lists the AI tool, the occupational classifications, and the number of positions no longer being filled. For a sourcer, that is a permanent negative signal: which job categories at which companies will never reopen.
Notices are filed with the EDD, the local workforce investment board, affected workers, AND city council members plus county supervisors in each jurisdiction. Multiple distribution points means multiple FOIA-able mirrors. Even if the EDD slow-walks publication, the same document lands on a dozen municipal agendas.
Penalties have teeth: back pay and benefits up to 60 days per worker, $500/day civil penalty, and a private right of action with attorneys' fees. There is also a Technological Displacement Act Fund seeded by those penalties, so enforcement gets its own budget. Translation: this dataset will get audited, and audit data tends to leak into the public record.
The vendor field is the actual unlock
Every existing tracker already produces names. WARNTracker.com, californiawarn.com, and layofflookout.com have been scraping the EDD's twice-weekly WARN Report for years. As of this writing, californiawarn.com shows 289 companies and 26,075 affected workers in the active 60-day window. Names are not new.
What SB 951 uniquely adds is the structured AI vendor disclosure. That is competitive intelligence wearing a recruiting hat. When Block (Jack Dorsey publicly attributed roughly 4,000 cuts to AI in a shareholder letter) files an SB 951 notice, you will know which specific vendor displaced which specific function. If it is Sierra eating tier-1 support, you pitch the displaced agents to Decagon. If it is Glean eating internal knowledge management, you pitch the displaced technical writers to a competing search vendor that still needs humans in the loop. If it is Cognition eating junior IC work, you reach the engineers before Anthropic's recruiting team does.
The same logic applies to Cloudflare's May 2026 cuts of 1,100, which Matthew Prince and Michelle Zatlyn framed around "the agentic AI era." Today that framing lives in a blog post. Under SB 951, it becomes a sworn government filing with a vendor named in a structured field.
The signal arrives before the badge, before the goodbye post, before the next hiring manager has written the JD.
The 90-day window breaks the existing playbook
Reactive sourcing has a fixed rhythm. News breaks, LinkedIn fills with "Open to Work" badges, every recruiter in the Bay Area runs the same Boolean string against the same people on the same day. Reply rates collapse. Senior engineers ghost. The good ones are gone in 72 hours, often to inbound from former coworkers who heard first.
SB 951 collapses that race entirely. The names are public 90 days before separation. The first-mover advantage shifts from "fastest LinkedIn search" to "fastest ingest of EDD filings, cross-referenced against GitHub, personal sites, and prior employers." Whoever wins the parsing pipeline wins the candidates.
This is exactly the kind of unstructured-to-structured problem we built Refolk for. You describe the person in plain English ("senior support engineers laid off from Block in the last quarter who previously worked on payments infrastructure, based in California") and Refolk pulls from GitHub, LinkedIn, and the open web to return a ranked shortlist. When the EDD starts publishing SB 951 filings, that becomes "senior support engineers at any employer that filed an SB 951 notice citing Sierra as the displacing vendor, with payments background." Same query shape. Better signal.
What you can already do, before SB 951 passes
The bill is not law yet. It cleared committee on June 10, 2026 and was re-referred. But two things make this worth wiring up now.
First, Governor Newsom's Executive Order N-6-26 (May 21, 2026) directs the EDD to launch a public AI-employment-impact dashboard by August 19, 2026, using UI data. It also orders the LWDA to recommend Cal-WARN revisions by November 17, 2026. Even if SB 951 stalls, the dashboard is coming. Newsom is not waiting for the legislature.
Second, the underlying pool is already enormous and already public. Roughly 150,000 tech layoffs in 2026 YTD. Challenger data attributes AI to between 17% and 26% of cuts. Named California-exposed AI-attributed cutters this year include:
- Cisco, around 4,000 (May 2026)
- Cloudflare, 1,100 (May 2026)
- Block, around 4,000
- Atlassian, around 1,600
- Coinbase, around 700
- Autodesk, around 1,000
- Oracle, 20,000 to 30,000
- Amazon, 16,000 corporate
That is more than 50,000 California-exposed workers from just three of those employers. Every one of them would have been a near-certain SB 951 filer.
Stanford's Digital Economy Lab adds an important wrinkle: employment for software developers aged 22 to 25 fell nearly 20% from its 2022 peak, while developers over 26 grew. The SB 951 dataset will disproportionately surface junior IC talent. If your req is "staff engineer with 10 years of experience," this dataset is not your friend. If your req is "sharp 24-year-old who shipped real code at a company that just got disrupted," it is the best lead source you will ever have.
The credibility filter nobody is talking about
Here is the contrarian read. NBER survey work found that 90% of executives said AI has had zero employment impact at their own companies, even as AI gets cited publicly as the reason for cuts. Most AI-attributed layoffs are AI-washing. The CEO needs a narrative for the shareholder letter, and "we are repositioning for the agentic AI era" sounds better than "we over-hired in 2022."
SB 951's required justification field changes the math. Putting "we replaced 400 support agents with Sierra" into a sworn government filing carries litigation risk that a blog post does not. The civil penalty is $500/day per violation, plus the private right of action. Some employers will quietly admit the cuts were not actually AI-driven and file under Cal-WARN instead. Others will file under SB 951 and have to defend the claim.
For sourcers, that creates a credibility-graded dataset. Workers laid off in a real, vendor-specified AI displacement are a different cohort than workers laid off in a generic restructuring. The former are domain experts in a function that just got automated by a specific tool. They know that tool's failure modes intimately. They are exactly who you want to hire for the team building the next version of that tool, or the team trying to compete with it.
This is where the natural-language layer matters more than the raw filing. A SQL query against EDD JSON gets you a list. Asking Refolk for "former Block support engineers who worked alongside Sierra deployments, now in market" gets you the cohort that competing vendors actually want, ranked, with paths to reach them. Same data. Different leverage.
How to wire this up before the bill passes
A few concrete moves.
Subscribe to the EDD WARN feed now. It publishes every Tuesday and Thursday. Treat it as your training set. When SB 951 enacts, the schema extends; the ingest pipeline does not change.
Build vendor watchlists, not company watchlists. If you recruit for Glean, your watchlist is "any company that files an SB 951 notice naming a competing enterprise-search AI." That list is empty today and will be long by Q2 2027.
Mirror the municipal agendas in your top three California metros. City council and county supervisor packets get posted before the EDD finishes processing. Quietly, this is where the leak path opens first.
Stop relying on "Open to Work." Stanford's data and LinkedIn's own denominator games already broke that signal. The next signal is structured, public, and 90 days early. Recruiters who treat the EDD as a primary source instead of a confirmation source will own the 2027 hiring market in California.
The bill might not pass this session. The dashboard is coming anyway. The trackers will mirror whatever the EDD publishes within weeks, the way they always have. The recruiters who win the next two years are the ones who set up the pipeline now and let the data fill it in.
FAQ
When does SB 951 actually take effect?
It does not yet. The bill passed committee 5-0 on June 10, 2026 and was re-referred to the Committee on P. & C.P. for further action. Even if it stalls, Newsom's Executive Order N-6-26 already directs the EDD to launch a public AI-employment-impact dashboard by August 19, 2026, and the LWDA to recommend Cal-WARN revisions by November 17, 2026. The data infrastructure is being built either way.
How is this different from existing WARN tracking?
Existing California WARN notices give you company, headcount, and separation date. SB 951 adds three structured fields that do not exist anywhere else: the specific job functions being automated, the AI system used, and the vendor that developed or sold it. That vendor field is the unlock. It lets you target displaced workers by which tool ate their job, which tells you which competing vendors will pay to hire them.
Will employers just file under Cal-WARN to avoid the AI disclosure?
Some will try. But SB 951's threshold is lower (25 workers or 25% of workforce, whichever is less, versus Cal-WARN's 50 workers), the penalty is steep ($500/day plus private right of action with attorneys' fees), and the parallel hiring disruption notice catches employers who quietly stop hiring for a role rather than running an explicit layoff. The bill was drafted with the workaround in mind.
What functions show up first in this dataset?
Based on 2026 AI-attributed layoffs to date, expect customer support, content moderation, technical writing, tier-1 sales operations, and junior software development. Stanford found developer employment for ages 22 to 25 fell nearly 20% from its 2022 peak. The SB 951 dataset will skew junior IC and operational, not senior engineering, at least initially.