SB 947 Passed 29-9. Your AI Screener's Real Problem Is the October FEHA Rule.
California's No Robo Bosses Act cleared the Senate 29-9, but FEHA's ADS rules are the live wire today. Here's what TA leaders should fix now.
The California Senate just sent the No Robo Bosses Act of 2026 to the Assembly on a 29-9 vote, and two days later Governor Newsom signed an executive order pointing the EDD and LWDA at AI-driven workforce displacement. Every TA leader I've talked to this week opened with the same question: "Does this kill our AI screener?" The short answer is no, not yet, and not for the reasons your legal team is forwarding around Slack.
The real exposure is already live. It's the FEHA automated decision system amendments that took effect October 1, 2025, and most recruiting orgs have not retrofit their pipelines for them. SB 947 and its WARN-flavored cousin SB 951 matter, but if you're optimizing for the wrong statute you'll spend Q3 hardening the layer that isn't on fire.
What actually passed, and what didn't
Sen. Jerry McNerney's SB 947, the No Robo Bosses Act of 2026, cleared the Senate 29-9 and is now in the Assembly. It would bar California employers from relying solely on automated decision-making systems (ADS) to discipline or fire workers, require independent human verification of any ADS-assisted termination, and prohibit ADS that use worker personal information to "predict" future behavior. Penalties run $500 per violation, with private right of action, attorney's fees, and potential punitive damages.
SB 947 is the post-veto rewrite of SB 7. Newsom vetoed SB 7 in October 2025 citing "unfocused notification requirements on any business using even the most innocuous tools." The rewrite narrowed scope. Importantly for sourcing teams, the provision that would have prohibited fully automated hiring was removed by the Senate Judiciary Committee with McNerney's consent. SB 947 primarily reaches discipline and termination. Hiring is adjacent exposure, not the headline prohibition.
Then there's SB 951, authored by Sen. Eloise Gómez Reyes, which is a separate bill people keep conflating with No Robo Bosses. SB 951 is a WARN-style notice law. It would amend Cal-WARN to require 90 days' notice (up from 60) for AI- or automation-attributed layoffs affecting 25 or more workers or 25 percent of a workforce, whichever is less. It would apply to employers with 75 or more employees, prohibit discharge without reasonable cause during the notice period at 100-plus headcount employers, and let displaced workers bid on other open roles. Its definition of "worker" sweeps in independent contractors employed for at least six months, which is broader than most WARN-type statutes.
The active compliance regime is FEHA, not SB 947
Here's the part the headlines bury. California's Civil Rights Council finalized FEHA amendments addressing automated decision systems in hiring, and they took effect October 1, 2025. That's the rule binding your screener today. SB 947 and SB 951 are future risk. FEHA ADS is present-tense liability, and disparate-impact claims under it don't need a new statute to fire.
If your ATS uses any of the following, you're in FEHA ADS territory: resume parsing that scores candidates, gamified assessments, automated video interview analysis, "culture fit" classifiers, knockout questions tuned by a model, or chatbot pre-screens that route candidates to "not advancing." Most teams have one or two of these and aren't sure which vendor owns the model. That ambiguity is the audit risk.
The cleanest fix is procedural. Document who the decision-maker is for every automated step, retain the inputs, retain the outputs, and make sure a human can re-derive the rejection rationale without the vendor's help. If a candidate files an FEHA complaint and you can't show the recruiter independently reviewed the screen, the rebuttable presumption flips against you.
Where Refolk fits, and where it doesn't
Sourcing tools and screening tools are not the same legal animal, and recruiters should stop letting vendors blur the line. Sourcing identifies candidates who might be worth a conversation. Screening makes adverse decisions about candidates who applied. FEHA ADS regulates the decision layer.
A tool that suggests "here are 40 backend engineers in Oakland who shipped to Postgres extensions in the last 18 months" is sourcing. A tool that auto-rejects 800 of 1,000 applicants based on a model score is screening. The first is largely outside the FEHA ADS perimeter. The second is squarely inside it, and SB 947's "predictive behavior" language reaches it again from a different angle.
This distinction is why we built Refolk the way we did. You describe the person you want in plain English, you get a ranked shortlist across GitHub, LinkedIn, and the open web, and the recruiter still owns every outreach decision. There is no auto-reject. There is no automated adverse action against an applicant. The model surfaces, the human decides.
The "human in the loop" standard is being defined upward
If you've been planning to satisfy SB 947 by adding a checkbox that says "recruiter reviewed," read the bill text again. SB 7's predecessor language required that any automated decision be reviewed by a person who must investigate and "compile corroborating or supporting information for the decision." SB 947 carries the same posture. Rubber-stamp review does not clear the bar. The reviewer has to independently corroborate.
For TA orgs, that means the days of "the model said no, recruiter clicked confirm" are numbered for any decision that touches a Californian. The realistic operating model is: model surfaces a recommendation, recruiter pulls independent evidence (work samples, references, recent GitHub or portfolio activity), and the rejection note documents the corroborating evidence rather than the model output. Anything less and the $500-per-violation math gets ugly fast at applicant volumes.
The model surfaces. The human decides. The note documents the evidence, not the score.
SB 951's vendor-naming requirement changes procurement
Tucked inside SB 951 is a clause that will reshape HR-tech contracts. The required notice content names "the AI system or automating technology that led to the displacement (including the entities that developed, sold, or leased the technology), and the justification and purpose for using the AI tool."
Translation: if SB 951 passes, your WARN notice will name the vendor. Vendors will demand sanitized language. Buyers will demand contractual indemnities. Procurement teams that handle HR tech should start adding AI indemnity riders to renewals now, because the vendors with the weakest legal answer will be the ones suddenly very busy renegotiating in Q4.
This is also why TA leaders should know which models touch their funnel. "We use the LinkedIn one" is not an answer. Neither is "the ATS has AI built in." If you can't list the systems by name, you can't write the notice, and you can't write the indemnity.
The EO is a 6-month regulatory ambush window
Newsom's Executive Order N-6-26, signed May 21, 2026, does not impose immediate obligations on private employers. It does build the apparatus that will. EDD is directed to launch an AI-impact dashboard using UI data within 90 days. EDD also reports twice yearly through 2027 on technology adoption's role in hiring and workforce decisions. LWDA has 180 days, roughly mid-November 2026, to recommend Cal-WARN modernization.
Anyone counting on Newsom vetoing SB 951 the way he vetoed SB 7 is misreading the politics. Independent trackers reported more than 114,000 tech-sector job losses across 150 companies in 2026, averaging roughly 825 per day, a pace 33 percent above the same period in 2025. TrueUp counted approximately 143,000 workers affected across 339 events. Roughly half of tracked layoffs this year have been explicitly attributed to AI by the companies making the cuts. That's the data EDD will be staring at when its first dashboard ships in August.
What to change before Newsom's desk gets a bill
Five concrete moves, in priority order.
1. Inventory every model touching your funnel
Name the system, the vendor, the input data, the output, and the decision it influences. If you can't produce this list in a week, you cannot comply with anything that passes. Refolk customers can ask in plain English for their own internal expert on a given model or vendor, which is faster than chasing procurement, but the inventory job itself isn't optional.
2. Separate sourcing from screening in your stack
Sourcing tools surface candidates. Screening tools make adverse decisions. Map every tool into one bucket. Anything that auto-rejects should be the highest-scrutiny system. Anything that only surfaces, like Refolk, should be classified accordingly so audits don't lump them together.
3. Rebuild the rejection workflow around independent corroboration
The recruiter's rejection note should reference the evidence they reviewed, not the model score. This is the SB 947 standard already in the bill text, and it's the defensible posture under FEHA today.
4. Add AI indemnity language to vendor renewals
If SB 951 passes in any form close to current draft, you'll be naming vendors in legal notices. Get the contractual coverage now while you still have leverage at renewal.
5. Stop hiring around the California problem
Some TA leaders are quietly routing California reqs to non-California recruiters. FEHA reaches the role, not the recruiter, and our internal data shows 63,567 U.S. TA leaders concentrated in NY and SF metros are already exposed because they hire into California roles even when based out of state. Geography arbitrage doesn't work here.
FAQ
Does SB 947 ban AI-based hiring decisions in California?
No. The fully automated hiring prohibition was removed from SB 7 by the Senate Judiciary Committee with the author's consent, and SB 947 inherits that narrower scope. The bill's express prohibitions reach discipline and termination, with adjacent exposure to hiring through the predictive behavior analysis clause. Hiring decisions made by ADS are regulated, but the binding rule today is FEHA, not SB 947.
What's the difference between SB 947 and SB 951?
SB 947 is the No Robo Bosses Act, authored by Sen. Jerry McNerney, requiring human review of ADS-assisted discipline and termination. SB 951 is the AI Job Killer Notice Act, authored by Sen. Eloise Gómez Reyes, requiring 90-day WARN-style notice for AI-attributed layoffs. They address different parts of the lifecycle. SB 947 passed the Senate 29-9. SB 951 is still working through committee at last public reporting.
Are sourcing tools regulated under FEHA ADS?
FEHA ADS regulates automated decisions about applicants and employees. Pure sourcing, where a tool surfaces candidates a recruiter then chooses to contact, is generally outside the adverse-action perimeter because no decision has been made about an applicant. Once a tool starts auto-rejecting or auto-routing applicants away from review, it crosses into ADS territory. The distinction matters and your vendors should be able to tell you which side of it their product sits on.
When should we expect SB 951 to land on Newsom's desk?
If it moves on a typical California timeline, late summer or early fall 2026 is the realistic window for floor passage in both chambers, with a veto-or-sign decision in the September-to-October range. Given the data flow from Newsom's EO and the political pressure described in the 114,000-job context, the veto path that worked for SB 7 in 2025 is a less safe assumption this cycle. Plan for it to pass.