Dark console-style card reading Was This Layoff Because of AI? Now They Have to Sign It

Was This Layoff Because of AI? Now They Have to Sign It

by
Reading Time: 5 minutes

Sach batao: if AI is behind a quarter of layoffs, why did not one New York filer say so on the official form?

AI layoffs are the loudest number in the job market right now. They may also be the least checked.

📈 The Number Everyone Quotes

Start with the headline figure, because it is real and it is big.

Challenger, Gray & Christmas tracks announced US job cuts and the reasons employers give. In its July 2026 report, AI was the top reason for the fifth month in a row: 10,970 cuts, or 33 percent of the month’s 33,429. So far this year, AI has been cited in 112,713 announced cuts, roughly 24 percent of all of them.

That is a lot of people, and a lot of headlines.

Here is the part that makes it sharper. Challenger has tracked AI as its own reason only since 2023, and it has logged 184,538 AI-cited cuts since then. Do the arithmetic and 112,713 of them, about six in ten, were announced in just the first seven months of 2026. The curve is steep, and it is why every executive, journalist and policymaker now has an opinion on it.

Now notice the verb. Challenger counts what employers cite. It does not audit what actually happened inside the company. Hold that thought.

🕳️ The Number Nobody Can Find

New York added a question to its mass-layoff (WARN) notice in March 2025: did technological innovation or automation play a part?

📌 Real-world case: In May 2026, the law firm Hunton Andrews Kurth reviewed the first year of those filings. More than 160 companies had filed. None had ticked the box. (Hunton)

So one record says AI is behind a quarter of cuts. Another, a legal form, says close to none.

Toh phir, which one is right? Possibly both, because they measure different things. The first is a press-release number. The second is a filing.

🏷️ What “AI-Attributed” Actually Means

Challenger is unusually honest about how fuzzy its own label is. In the July report, it walks through two cases.

📌 Real-world case: Visa announced a 7 percent reduction and described it as an efficiency push in which AI would reshape work. Challenger filed it under AI. Separately, the Bronx hospital system Montefiore cut 12 utilization-review nursing positions after adopting software from Datavant. A nurses’ union called it replacing humans with AI; hospital leaders called that characterisation misleading. With no clarity on the product, Challenger logged it as “Technological Update (possibly AI)”. (Challenger, July 2026 report)

Read those two together. One company volunteered AI as the story. The other had its cuts argued into the category by a union and a spreadsheet.

Same label, very different evidence.

Challenger even keeps a separate bucket for the grey zone. It uses “Technological Update (possibly AI)” when a company says new technology was the main reason for cuts and AI is alluded to but never directly tied to them. That bucket held 20,219 cuts in 2025. In other words, a meaningful slice of the story sits in a category whose own name is a question mark.

⚖️ Two Ways to Be Wrong

Once you see the fuzziness, you can see two opposite errors.

Overcounting. A company with a different problem, such as over-hiring or weak demand, says “AI” because it sounds forward-looking. The cut is real. The cause is borrowed. This is the dressing-up worry Oxford raises.

Undercounting. A company genuinely replaces tasks with software but says “restructuring” or “efficiency” because naming AI would cost it goodwill with staff. The cause is real. The label is missing.

A single total, like 112,713, quietly mixes both errors. You cannot tell from the number alone which one is bigger.

That is not a reason to ignore the data. It is a reason to ask what each row is really measuring before building a policy, a career plan or a panic on it.

🎭 Why Say “AI” in a Press Release but Not on a Form?

This part is interpretation, so treat it as my reading, not a proven fact.

A press release is a story told to investors. A WARN filing is a record kept by a government. Challenger itself notes that naming AI in a layoff announcement can win over investors while pushing current and prospective employees away, and that messaging has swung from hedging to citing it aggressively.

An outside check points the same way. Oxford Economics, in a January 2026 briefing reported by Fortune, found that AI was blamed for roughly 55,000 US cuts in the first 11 months of 2025, about 4.5 percent of the total, against 245,000 tied to market and economic conditions. Oxford suspects some firms are dressing up layoffs as a good-news story. It also pointed out that if AI were replacing labour at scale, productivity growth should be speeding up, and generally it is not. (Fortune)

Hunton gives two fair explanations for the empty New York box. Either AI is not really driving layoffs yet, or the form is unclear. The statute was not amended, so employers may not know whether the question is optional, how to define automation, or whether AI must be the only cause.

Notice that neither explanation says companies are lying. The ambiguity does the work.

✍️ Connecticut Turns It Into a Signature

That is why the next rule matters.

Since 1 October 2026, Connecticut requires employers whose layoffs trigger a WARN notice to tell the state Department of Labor, in writing, whether the layoffs relate to their use of AI or other technological change. It applies whenever a mass layoff or plant closing triggers a notice, not only when AI picked who got laid off. It is part of a wider AI law, SB 5, signed in May 2026, and the state attorney general has sole enforcement authority over the law’s notice requirements. (Epstein Becker Green)

I could not find a penalty specific to this disclosure, so I will not invent one. And I have seen no filings under the new rule yet, so I cannot tell you what Connecticut’s numbers will look like.

There is a catch worth flagging. Challenger’s own view is that as regulations take shape, companies will be more careful in how they announce cuts, which could make tracking AI’s real effect on jobs more opaque, not less.

A disclosure rule can produce honesty. It can also produce better wording.

🪙 Token Truth

A label is not a measurement.

A count of “AI layoffs” is a count of announcements and the reasons attached to them. It is not a count of jobs a model actually did. That is the same trap this blog keeps meeting with tokens: easy-to-count numbers get treated as proof of value. (If you want that thread, start with Episode 1 of The Token Diaries.)

🔎 How to Read the Next AI-Layoff Headline

Seedhi baat: three quick checks.

  1. Who is saying it, and where? A press release and a legal filing carry different stakes. Look for the filing.
  2. What is the evidence behind the label? Did the company name a product and a process, or just “efficiency”? Visa’s wording and Montefiore’s are not the same kind of claim.
  3. What else was happening? Over-hiring, weak demand and cost pressure all end in layoffs too. In 2025, Oxford found, they outnumbered the AI-blamed ones roughly four to one.

None of this says AI is not changing work. It is. The question is how much of this layoff, at this company, you can actually attribute to it.

👀 What I Am Watching

Connecticut’s first filings, and whether New York’s box ever gets ticked.

If the new forms start showing AI as a stated cause, the headline numbers will have something real to be checked against. If they stay empty while press releases keep naming AI, that gap is the story.

If your employer had to sign a form saying whether AI caused the layoffs, do you think the answer would change what they say in public?

📚 References

What’s your Reaction?
+1
0
+1
0
+1
0
+1
0
+1
0
+1
0
+1
0