Connecticut's AI Layoff Disclosure: The WARN Question and New York's Zero Yeses

Since 1 October 2026, every employer filing a federal WARN notice with Connecticut's Department of Labor faces a new question. It must say whether the layoffs relate to its use of artificial intelligence or another technological change. Connecticut's AI layoff disclosure is the first such question written into a state statute and already in force. It is not the first time a state has asked. New York asked more than 160 companies a similar question and recorded 0 yes answers.

The short answer. If your company has 100 or more employees and files a WARN notice in Connecticut, the rule applies to you. You must now tell the Labor Department whether the layoff relates to your use of AI or another technological change. No official form exists yet. No penalty is attached to this sentence, and the state has not said it will publish the answers.

What Connecticut's AI layoff disclosure actually requires

The provision sits inside Public Act 26-15, the state's broad AI law that began life as Substitute Senate Bill 5. Governor Ned Lamont signed it on 29 May 2026, according to Ogletree. Nixon Peabody described the full text as a 74-page act, and almost all of it concerns things other than layoffs.

The layoff part is a single obligation. It applies to each employer that serves written notice on the Labor Department under 29 USC 2102(a), the federal WARN notice section. That employer must disclose "whether the layoffs that are the subject of such written notice are related to the employer's use of artificial intelligence or another technological change". The disclosure is made "in a form and manner prescribed by the Labor Commissioner".

That is the whole rule. It adds a line to a filing that already existed. It creates no new threshold, no new waiting period and no new right for workers to see the answer.

Who it catches

The trigger is the federal WARN Act, not a separate Connecticut test. Federal WARN generally covers employers with 100 or more employees and requires 60 days of notice before a plant closing or mass layoff. If you are not filing federal WARN notice in Connecticut, this provision does not reach you.

The scope of the question is wider than many summaries suggest. Epstein Becker Green's reading is that the duty "is not limited to disclosing the use of AI or other technology to make layoff selections". Proskauer frames it the same way. The question is about cause, so a team cut because an agent now does its work is squarely inside it.

What it leaves undefined

"Related to" is the load-bearing phrase, and the act does not define it. A layoff where AI did 10% of the reasoning and one where it did all of it read identically on the face of the statute. Your answer depends on your own reading.

"Another technological change" is also undefined, as far as published commentary shows. That matters because almost every restructuring involves new software somewhere. A literal reading could cover a migration to a new ERP system as easily as a support agent rollout.

Artificial intelligence itself is defined by cross-reference to another section of the act. The definition follows the now-common machine-based system formula that infers from inputs how to generate outputs. That is broad enough to include tools most HR teams would not call AI.

The Connecticut WARN form that does not exist yet

As of 12 October 2026, the Department of Labor's WARN guidance page does not mention artificial intelligence, technological change or Public Act 26-15. We read it on that date. It still describes a WARN notice as a letter on company letterhead, signed and dated, emailed or mailed to the Rapid Response Unit in Wethersfield.

The page lists what that letter should contain: the affected site, the dates, the number of workers and an employer contact. It carries no field, checkbox or template for the new disclosure. The guidance page itself is dated August 2025, which predates the law.

This is the gap employers are filing into. The statute requires a disclosure in a form the Commissioner prescribes. The Commissioner has not visibly prescribed one. Until that changes, the safest course is a clearly labelled paragraph in the notice letter that answers the statutory question in the statute's own words.

What the Labor Commissioner has to decide

Three design choices will decide whether this produces useful data or noise. The first is whether the answer is a binary yes or no, or a structured field with the share of roles affected. The second is whether "related to" gets a threshold, such as primary cause or substantial factor.

The third is publication. The statute requires disclosure to the department, not to the public. Connecticut publishes a list of WARN notices received. Nothing in the commentary we read says the AI answer will appear on it. If the answers stay inside the agency, this is a record, not a disclosure in any meaningful public sense.

My view is that the Commissioner should copy California's field list rather than New York's checkbox. The statute gives enough latitude to ask for numbers. A form is also far easier to change than a law.

Connecticut is not the first state to ask, and New York shows why that matters

The framing that Connecticut created the first official record of AI layoff claims in the US is wrong. The correction is useful. New York got there 18 months earlier, by changing a form instead of a law.

In January 2025 Governor Kathy Hochul directed the New York State Department of Labor to gather more data on layoffs. The department added an AI-related disclosure to its WARN filing system in March 2025. Employers are asked whether "technological innovation or automation" contributed to the layoffs, by marking a checkbox. A box that is ticked opens a follow-up asking the employer to name the technology.

The statute behind New York's WARN Act was never amended to require this. Hunton Andrews Kurth noted that as a result, employers "may question if the disclosure is optional or mandatory". Connecticut's version removes that doubt. That is its single real improvement on New York.

160 companies, zero yeses

The New York result is the most important number in this whole debate. By May 2026, more than 160 different companies had filed WARN notices with the New York DOL, and not one attributed layoffs to AI or automation. The reasons they gave instead were economic circumstances, contract terminations, bankruptcy, sale of the business and relocation.

Hunton put the two possible readings plainly. Either AI is not contributing to mass layoffs in New York, or it is and businesses are not reporting it. Both readings are uncomfortable for anyone who wants a self-reported checkbox to settle the question.

Compare that with the private counts. Challenger, Gray and Christmas tallied more than 17,000 job cuts tied directly to AI in the first 9 months of 2025, as reported by CIO Dive. Those figures come from what companies say in press releases and earnings calls. When the same question arrives on a government form, the answer goes quiet. That gap is the story. Our earlier analysis of how companies word AI layoff statements covers the press release side of it.

Three designs for AI layoff reporting, compared

Put the three state approaches side by side and the design choices become obvious. Each asks a different question, sends the answer to a different audience and attaches a different consequence.

California's version names the technology

California's SB 951, signed by Governor Gavin Newsom on 30 September 2026, goes much further than Connecticut. A Cal/WARN notice for a technology displacement must state the number of layoffs substantially attributable to AI or other automated technology. It must also list the occupations and work locations involved, and the job functions that will be automated.

The notice must name the specific category of AI system or automating technology. It must also open with the sentence "This notice is for a technology displacement". The Employment Development Department will post a summary of the notices it receives and a quarterly statewide summary, according to Ogletree's October 2026 alert.

That is a disclosure regime. Connecticut's is a survey question. If you sell software into HR or operations teams, California's rule is the one that changes what buyers ask you. That point also runs through the breakdown of California's AI laws for software vendors.

Other states are moving too. MultiState counted proposed WARN amendments for AI in Pennsylvania, Illinois and Minnesota as of August 2026. It also tracked a New Jersey bill requiring 90 days' notice for AI-driven layoffs. At federal level, Senators Josh Hawley and Mark Warner announced the AI-Related Job Impacts Clarity Act in November 2025. It would require quarterly AI layoff reports from public companies and federal agencies. It has not become law.

The running count of Connecticut AI layoff disclosures

This post was meant to carry a running count of filings made under the new rule. Here is the honest state of that count on 12 October 2026, eleven days after the rule began.

This table will be updated as the Labor Department publishes a form or adds a field to its list. A zero here means no answer is publicly visible, not that every employer answered no.

The scale is small regardless. Connecticut employers filed about 13 WARN notices in 2026 through early August, according to Polauf Law. Of those, 8 covered more than 50 workers. At that pace, the state will see perhaps a handful of filings a quarter under the new rule.

Why the count may stay near zero

Every incentive points toward "no". A yes invites press attention, union questions and, in some readings, litigation risk under the act's separate rules on automated employment decisions. A no costs nothing, because no penalty attaches to this disclosure in any of the law firm alerts we read.

Nixon Peabody, Proskauer and Ogletree all describe Attorney General enforcement for the act's automated decision notices. That enforcement carries a cure period running to 31 December 2027. None of them extends that enforcement to the WARN sentence. A rule with no consequence for a wrong answer tends to produce the answer that is safest for the filer.

That matches a pattern already visible in disclosures elsewhere. The SEC has gone after companies for overstating AI in investor materials, as logged in the record of AI washing enforcement cases. Connecticut's form invites the opposite error: understating AI to the regulator while overstating it to investors.

Where this argument is weakest

The case above treats a yes-or-no question with no penalty as close to useless. That case has two real problems.

A mandatory question changes internal behaviour even if the answer is no

New York's question was optional in practice. Connecticut's is not. A legal duty to answer means someone inside the company has to decide, before filing, whether AI related to the layoff. That decision leaves a document trail, and a document trail is discoverable in later employment litigation.

So the rule may change what companies record more than what they report. That is a meaningful effect. The zero count in a public tracker cannot measure it. I think this is the strongest argument for the law. It is not one its supporters usually make.

New York's zero may be partly true

It is possible that very few mass layoffs at the WARN threshold are driven by AI yet. Much of the AI attribution in private counts comes from technology and services companies cutting white-collar roles in small tranches. Those cuts may never trigger WARN at a single site. Our analysis of what payroll data shows about AI job losses found the macro signal weaker than the headlines.

If that is right, a zero on the form is accurate and the instrument is fine. It is pointed at a population that has not yet been hit. The debate over whether many AI layoffs are really a correction of pandemic overhiring remains open. I still lean toward under-reporting, because companies talk about AI efficiency loudly everywhere except on government forms. But the evidence for that is circumstantial.

What to record before you file a Connecticut WARN notice

If you run operations or HR at a company with Connecticut sites, the practical work is small and worth doing now. The question will be asked, and the answer will be read later by people who were not in the room.

First, decide what "related to" means inside your company and write it down. A sensible internal test is whether AI or new technology removed or materially reduced the work of the affected roles. It is not whether any software changed. Apply it the same way to every filing.

Second, keep the reasoning, not just the answer. Record which roles were affected, what work they did, and whether that work is now done by a tool, by other staff or not at all. If California's rule later applies to you, that record is most of what its notice requires.

Third, check your public statements against your filing. If an earnings call or all-hands credited AI efficiency for the same reduction, a "no" to the state is hard to defend. I would rather answer yes with a short explanation than answer no and contradict my own press release.

Frequently asked questions

Where to start this week

Write one paragraph that defines "related to AI or another technological change" for your company. Get legal and HR to sign the same version. Every future WARN letter in Connecticut, and every Cal/WARN notice from January, will lean on it.

Then pull your last two reductions in force and answer the question for each as if you were filing today. If the honest answer differs from what you said publicly at the time, you have found the problem before a regulator or a plaintiff's lawyer does.

Related reading