Connecticut's CART Act Adds an AI Checkbox to Every WARN Notice

On October 1, 2026, Connecticut quietly became the first US state where every WARN notice has to answer one new question in writing: was this layoff "related to the employer's use of artificial intelligence or another technological change?" The first filings carrying that disclosure land this month. For anyone whose job is finding displaced technical talent fast, this is the cleanest signal the US public record has ever produced.

What the CART Act actually changed on October 1

Connecticut's Public Act 26-15, the AI Responsibility and Transparency Act (CART Act), now requires every employer filing a Connecticut WARN notice to disclose whether the mass layoff is related to the employer's use of AI or another technological change. It is the first US state WARN regime with an explicit AI-causation checkbox, and the disclosure provision took effect October 1, 2026.

Three things to internalize before you build a workflow around it:

• The flag attaches to any federal WARN filed with the Connecticut Department of Labor. Federal WARN kicks in at 100 employees and requires 60 days written notice.
• Connecticut is the sixth US state to regulate how employers use AI in hiring and workforce decisions, but the first with an AI-causation WARN flag.
• Most of the compliance work employers have been doing - interaction disclosures and pre-decision notices for automated hiring tools - is aimed at obligations that do not take effect until October 1, 2027. The WARN flag landed with employers unprepared.

The practical read: the October and November 2026 filing waves are the first batch where the "AI" box is live, and employment counsel at firms like Epstein Becker Green and Ropes & Gray are still writing the interpretive guidance in real time.

The baseline: Connecticut WARN is small, and that is the point

Connecticut is a thin market, which is exactly why the AI flag matters so much. A single 100-person AI-flagged event is a measurable fraction of the state's visible technical workforce.

Here is the dataset to anchor on:

0.006%

The 42% year-over-year jump in affected workers already made 2026 an unusual year before the flag went live. Layer the AI disclosure on top and you have a feed where every row carries both a reason code and a 60-day countdown.

Why the AI flag will be noisier than you expect, in a useful way

Treat the AI flag as "tech-adjacent," not "replaced by GPT." Risk-averse employment counsel will mark "yes" defensively on borderline events, which inflates the AI-flagged subset: good for sourcing volume, mixed for signal purity.

The reason is baked into the statute. The disclosure forces judgment calls about whether automation, AI adoption, or technology modernization contributed to a reduction in force. If you are a general counsel staring at an ERP migration, an RPA rollout, or an offshoring plan with "AI-assisted ops" in the project charter, you check the box. The downside of over-flagging is a strongly worded letter. The downside of under-flagging, once CT courts read the Act in concert with the Connecticut Fair Employment Practices Act, is substantial potential liability.

That cuts two ways for sourcers:

1. The "AI-flagged" subset will capture plain-vanilla tech modernization (SAP cutovers, Workday rollouts, call-center automation), not just LLM displacement. Widen your title lens accordingly.
2. There will be a counter-trend of under-flagging at employers whose counsel reads PA 26-15 as a liability trap. Scrape the free-text sections too; the word "automation" in a narrative paragraph is as actionable as the checkbox itself.

The AI flag is a sourcing primitive, not a verdict. The checkbox tells you where to start reading, not who to call.

The real poach list is Hartford insurance, not Stamford tech

The AI-flagged WARN subset in Connecticut will skew toward actuarial, claims, underwriting, and clinical-ops roles, not laid-off software engineers. Technology was 1 notice and 2 employees across the last 24 months in CT. Finance, insurance, and healthcare dominate.

Industry mix over the last 24 months:

• Manufacturing: 10 notices, 1,230 employees
• Healthcare: 10 notices, 896 employees
• Retail: 4 notices, 1,215 employees
• Finance & Insurance: 3 notices, 328 employees
• Technology: 1 notice, 2 employees

The non-obvious implication: the first wave of CART-flagged WARNs will hand insurtech, healthtech, and AI-ops startups a pre-qualified list of domain SMEs whose jobs were just automated away. These are the hires that are hardest to make from a cold search, because the titles ("Senior Underwriting Consultant," "Clinical Informatics Lead," "Claims Operations Analyst") do not Boolean cleanly. You need a tool that reads intent, not keywords, which is the exact gap Refolk closes: you describe the person in plain English ("ex-Aetna underwriters in Hartford who worked on claims automation") and get a ranked shortlist back without hand-tuning strings.

Named filings to watch for refilings or follow-on waves after October 1:

• CVS Health / Aetna (Hartford): 313 employees assigned to Aetna's Hartford office, including 17 Connecticut residents, laid off between April 3 and July 31.
• Talcott Resolution Life (Hartford): 101-worker closure. Life-insurance back-office is a canonical GenAI target.
• Stanley Black & Decker (New Britain): 300-worker closure. Manufacturing plus ERP and automation modernization.
• IDEX Health & Science (Bristol): 73 workers.
• JeniusBank (Hartford): 161-worker closure filed January 8, 2026. Digital-bank wind-down, likely tech-heavy headcount.

Any of these that file a supplemental or follow-on notice this fall will be the first employers in the US to carry a formal "AI contributed to this layoff" state-government disclosure. That is an outreach line you cannot buy.

The CHRO disclosure cascade: a second, better signal

The best personalization hook in the CART Act is not the checkbox. It is the follow-on disclosure the Connecticut Department of Labor is empowered to pass to affected employees.

The Act contemplates that the DOL first receives notice that a layoff was related to AI or another technological change. Individual employees can then receive information identifying the technology used, its purpose, the categories and sources of personal data processed, and the manner in which those data were assessed. In plain English: the worker who just got laid off is handed a document that names the system that replaced them.

For sourcers, that is a two-step cadence:

1. The WARN flag goes public. You know an employer did an AI-related layoff. You start a watchlist.
2. The named-system disclosure reaches the employee. You get a personalization vector no competitor has: "I saw your CART disclosure named [vendor X]. The team at [startup] is building the exact counterweight to it."

Nothing else in the US WARN regime currently produces an artifact like that. If you are running outbound into displaced talent, the one-line opener that references the system by name will outperform anything generic by a wide margin.

The 100-employee floor is where most AI displacement actually hides

The CART Act flag only surfaces 100-plus-headcount events, so it will miss roughly 80% of the real AI-driven workforce reduction happening in Connecticut right now. Build your workflow for both.

Federal WARN reaches employers with 100 or more employees. Connecticut adds no state requirement, so a layoff below those thresholds is entirely lawful with no notice at all. That means the kind of displacement that actually tracks with generative AI (five-person content teams quietly cut, eight-person support pods replaced by a Zendesk AI tier, a dozen junior designers let go when a Figma plugin lands) will never trigger a CART flag.

Three things to layer on top of the WARN feed to close that gap:

• Follow Connecticut-based GitHub contributors whose commit cadence falls off a cliff. Mid-career engineers rarely stop pushing on purpose.
• Track LinkedIn "Open to Work" badges filtered to Hartford, Stamford, New Haven, and Fairfield County. The AI-flagged WARN crowd will light these up within two weeks of their 60-day clock starting.
• Pull alumni of the known filers (Aetna, Talcott, Stanley Black & Decker, JeniusBank) who left in the 90 days before the formal notice. Many quiet exits precede a WARN by one quarter.

How to operationalize the flag before Thanksgiving

Treat October through December 2026 as the uncontested window. The flag is new, the compliance bar is still being set, and most recruiting teams will not have built a filter for it yet.

A workable playbook for the next 60 days:

1. Build a daily scrape of the Connecticut DOL WARN page, warnact.io, layoffdata.com, warnfirehose.com, and kadoa.com/layoffs/state/CT. Parse both the structured flag and the narrative sections for "AI," "automation," "machine learning," and "technology modernization."
2. For every flagged filing, pull the employer's last 12 months of job postings. The roles they were hiring for right before the WARN tell you exactly which functions got consolidated.
3. Cross-reference flagged employers against your ATS. Silver-medal candidates from Aetna, Talcott Resolution, Stanley Black & Decker, and JeniusBank who passed on an offer in the last two years are warm again.
4. Use a plain-English search tool like Refolk to pull alumni cohorts across LinkedIn, GitHub, and the open web without hand-building Boolean strings for every title variant.
5. Draft two outreach templates: one that references the CART flag directly ("I saw your employer disclosed AI as a factor in the recent WARN"), one that stays generic for employees who have not yet received the named-system letter.

2,371

The CART Act did not create new layoffs. It created a public, structured, filter-ready reason code on top of a flow that was already running 42% hotter than last year. The sourcers who build the filter this month get a six-to-eight-week head start on everyone who waits for the first conference panel to explain it.

FAQ

Does the CART Act require employers to disclose which AI vendor caused the layoff?

The WARN-level flag itself only asks whether the layoff is related to the employer's use of AI or another technological change. The named-system detail (what technology was used, its purpose, the data it processed, and how it was assessed) sits in a follow-on disclosure that the Connecticut Department of Labor can route to individual affected employees. In practice, that means sourcers see the flag first, and workers see the vendor name shortly after, which is the window where a well-timed personalized outreach lands best.

Will the AI flag be reliable, or will employers under-report?

Expect over-reporting in the first few quarters, not under-reporting. Risk-averse employment counsel will check the box defensively on anything tech-adjacent, including ERP migrations and process automation that have nothing to do with generative AI. There will be a counter-trend of employers who read PA 26-15 in concert with CFEPA as a liability trap and quietly under-flag, so always read the free-text narrative, not just the structured field.

How many engineers are actually in Connecticut to poach?

In Refolk's index of professional profiles, exactly 40 US engineers with Software, Senior, Staff, ML, Data, or Engineering Manager titles are tagged "Connecticut," against 633,674 nationwide in the same cut. Top employers in the CT set include General Dynamics Electric Boat, Cognizant, Veeder-Root, Gerber Technology (Lectra), Timex Group, BlackRock, Belcan, and TikTok USDS JV. There are zero ML Engineers in CT metros tagged with Machine Learning skill, which tells you most displaced AI-adjacent talent will relocate or go remote.

What sectors will the AI-flagged WARNs actually come from?

Finance, insurance, and healthcare, not technology. Over the last 24 months, those two sectors accounted for roughly 39% of CT WARN notices, while technology filed one notice covering two employees. The practical poach list will be dominated by Hartford-area insurance roles (actuarial, underwriting, claims operations) and clinical-operations roles at hospital systems, which is exactly the SME profile insurtech, healthtech, and AI-ops startups struggle to source cold.