AI Fell to #5 in September. SB 951 Explains Why.

Challenger's September 2026 report, released October 1, showed something strange. After five straight months of leading the chart, "AI" fell to the #5 cited reason for U.S. job cuts, with just 3,961 attributed to it (about 9% of the month). The same week, Governor Newsom signed SB 951, which forces California employers to label AI-driven layoffs starting January 1, 2027. The two events are not a coincidence, and the gap they open is the one your resume has to work inside.

What actually happened in September 2026

AI did not stop replacing workers in September. Employers stopped saying so out loud. Challenger counted 43,281 U.S. cuts, down 18% from August and the lowest September total since 2022, and the share of those cuts attributed to AI collapsed from 33% in July to 9% in September.

The numbers that moved:

• AI cuts, September 2026: 3,961, down from leading the chart for the prior five months
• AI cuts, year to date 2026: 120,136, roughly 21% of the 573,195 YTD total
• New #1 cited reason in September: "market and economic conditions," at 8,789 cuts
• Tech sector, YTD: 165,925 cuts, about 29% of all 2026 layoffs
• California, YTD: 114,818 cuts, leading the nation

AI remains the biggest single named driver of 2026 layoffs. It just stopped being the biggest driver of September press releases. Andy Challenger, the firm's chief revenue officer, said naming AI can "win over investors while pushing current and prospective employees away," and warned regulation would make "tracking the impact of AI on jobs more opaque." He said that before SB 951 was signed. The September report is him being right in real time.

What SB 951 actually does, in one page

SB 951 amends California's WARN Act (Labor Code §§1400.5, 1401, 1402.7) to require Cal/WARN notices to disclose when a layoff was "caused in whole or in substantial part by" AI or automated technology. Governor Newsom signed it September 30, 2026. It takes effect January 1, 2027.

The notice itself has to carry a specific top line: "This notice is for a technology displacement." Beyond that, employers must disclose:

1. Which roles are affected
2. Which job functions were automated
3. The type of AI technology deployed

California's Employment Development Department (EDD) will publish quarterly statewide summaries of technology-displacement notices, and must report to the Legislature by January 1, 2028 on AI's effect on hiring. A sister bill, SB 947 ("No Robo Bosses Act," effective July 1, 2027), bars sole reliance on an automated decision system for discipline or firing. Senator Eloise Gómez Reyes authored SB 951.

The loophole SB 951 does not close

SB 951 only triggers on Cal/WARN thresholds: 75+ employees at the establishment, 50+ affected. Sub-threshold cuts, contractor terminations, and (most importantly) roles that quietly get frozen and never re-posted are entirely outside its scope. The EDD's own 2028 report has to look at hiring freezes caused by AI precisely because the law does not.

If you have been applying into a market that feels thinner than the layoff numbers suggest, this is why. The jobs most likely to be delabeled or never re-posted are the exact ones most exposed to automation, and most dispersed across the country.

The 30 to 1 ratio nobody is talking about

In Refolk's index of professional profiles, there are roughly 30 customer support workers in the U.S. for every AI or ML engineer. There are nearly 4 copywriters and content marketers for every one. The retraining math does not work at the population level, and that fact drives everything about how you should rewrite your resume.

The top employer in the support cohort in Refolk's index is Intuit TurboTax Verified Pro, a contractor-model operation that sits outside Cal/WARN's 75-employee establishment trigger entirely. The AI-builder pool clusters at Meta, Distyl, Sandgarden, and Scout Space, concentrated in the Bay Area, NYC, and Seattle. The exposed pools are dispersed everywhere else.

Even if every AI engineer job in the U.S. doubled overnight, it would absorb less than 7% of the support pool. "Become an AI engineer" is not a strategy for the median exposed worker. "Become the human in an AI-augmented workflow" is.

Why "market and economic conditions" is the phrase to watch

September's new #1 cited reason, "market and economic conditions" at 8,789 cuts, is the laundering category. It is broad enough to absorb automation decisions without triggering either investor-facing AI narratives or Cal/WARN disclosure obligations, and you should expect to see it in a lot more Q4 2026 memos.

Challenger has already created a sub-category called "Technological Update (possibly AI)" for cuts where AI is alluded to but not named. Their July 2026 Datavant case study, where AI was "alluded to but not directly tied" to a hospital-system cut, is the prototype. SB 951 is designed to close that euphemism loophole in California. Everywhere else, the euphemism is going to get worse before it gets better.

The practical consequence for a job seeker: the layoff notice you were part of, or the one your target team went through last quarter, probably does not say "AI" even if automation was the proximate cause. Your resume has to translate the work anyway.

How to write a resume for a role that was quietly automated

Describe the work in terms the next employer, who is also deploying AI, wants to buy: lead with judgment, exception handling, and measurable throughput, not with the tool you used to do the task.

Five rewrites that actually move interviews:

1. Replace tool nouns with outcome verbs. "Handled 80 Zendesk tickets per day" becomes "resolved 80 customer issues per day at a 92% first-contact rate, including 12 that required policy exceptions."
2. Name the AI you already worked alongside. If you QA'd an LLM's drafts, triaged chatbot escalations, or trained a model with your corrections, say so. "Human in the loop for a GPT-4 class support agent, reviewed 1,200 drafts/week, caught 7% hallucination rate on refund policy."
3. Quantify judgment calls, not volume. Volume is what gets automated first. The exceptions, the escalations, and the policy edges are what stay human. Count those.
4. Translate "content writer" into "editorial judgment." Brief development, voice calibration across 6 brands, legal review coordination, SME interviews. The parts a model cannot do alone.
5. Lead with the function, not the title. "Customer Support Specialist" is a title an ATS filter can downrank in 2026. "Policy exception analyst, Tier 2 escalations, insurance claims" is the same job described in a way a hiring manager scanning for human-in-the-loop talent will stop on.

Translating two years of ticket volume into exception-handling language is the exact work Refolk takes off you: paste the posting, and Refolk writes a resume from your own history tailored to that specific role, and scores how well you actually fit before you send it. For a support or content pro repositioning against delabeled automation cuts, that fit score tells you whether the translation landed or whether you need to pull more specifics from your last two years of work.

The California angle, if you are job hunting there now

California led the nation in 2026 layoffs with 114,818 YTD cuts through September, and starting January 1, 2027, Cal/WARN notices for AI-driven cuts in the state will carry the specific disclosure "This notice is for a technology displacement." If you are applying in California over the next 90 days, three things to do:

• Watch quarterly EDD summaries starting in Q1 2027. They will name roles and job functions automated, by employer. That list is the best public map of which teams are actively replacing people with software and which are backfilling.
• Check if your last WARN notice qualifies for retroactive context. It will not be amended, but if it was filed in late 2026 and your role matches a function that shows up in the first EDD summary, your resume can cite the function (not the employer) accurately.
• Assume anything cited as "market and economic conditions" at a California tech employer in Q4 2026 is a candidate for the technology-displacement bucket. You will not be able to prove it. You can still frame your experience as human-in-the-loop ahead of the hiring manager figuring it out.

A University of Pittsburgh study from May 2026 found productivity did not actually rise at firms that invested in AI and cut staff. That is the counter-narrative your cover letter can lean on when a hiring manager is still deciding whether to re-hire the function.

What the NY Fed number tells you about your odds

The New York Fed's September 2026 survey found only 4% of NY-area service firms reported laying off staff due to AI, against Challenger's 21% YTD cited share. That gap is the real story.

Employers are more willing to tell investors and the press that AI drove a cut than to tell a regulator or a survey-taker. SB 951 is the first law that forces those two numbers to converge, at least in California. For the next 15 months, the honest signal about what is being automated will come from three sources:

• Quarterly EDD technology-displacement summaries (first drop expected Q1 or Q2 2027)
• Challenger's "Technological Update (possibly AI)" sub-category trend line
• Job postings that describe roles as "AI-augmented," "human review," or "escalation only"

Treat the third one as a tell. Postings that use that language are the ones where the function was partially automated and the remaining human seat is the one you want.

FAQ

Is AI actually slowing down as a cause of layoffs?

Almost certainly not. Challenger's own CRO predicted employers would get more careful with language once regulation hit, and SB 951 was signed September 30, the same month AI dropped from the #1 to the #5 cited reason. Tech sector cuts stayed at roughly 29% of the YTD total. The underlying rate of automation-driven cuts is not visible in the September headline number, which is why the "market and economic conditions" category jumped to the top.

Does SB 951 apply to my employer if we are outside California?

No. SB 951 amends California's WARN Act and only governs notices filed with California's EDD for California layoffs at establishments of 75 or more employees affecting 50 or more. If you were laid off elsewhere, there is no federal equivalent yet, and the EDD's quarterly summaries will only cover California. That is still useful intel, because large multi-state employers will often automate the same function in multiple states, and the California disclosure will tell you which function.

Should I put "AI" or "LLM" on my resume if I used them at my last job?

Yes, if you can describe what you did with judgment, not just use. "Used ChatGPT" is noise. "Reviewed 1,200 LLM-drafted support replies per week, flagged a 7% hallucination rate on refund policy, retrained prompt library quarterly" is signal. Hiring managers in late 2026 are explicitly looking for people who have worked alongside AI without being either threatened by it or oversold on it, and specifics are how they tell the difference.

How do I know if a role was quietly automated before I apply?

Three tells: the posting is a re-opening of a role closed six to twelve months earlier at a smaller headcount, the job description includes language like "escalation," "review," "oversight," or "exception handling," and the team structure in the posting shows one senior role replacing what used to be a pod. None of these are proof. All three in one posting is a strong signal that the function was partially automated and the role you are applying for is the human-in-the-loop seat.