Starting October 1, 2026, any Connecticut employer filing a WARN Act layoff notice must also tell the state Department of Labor whether artificial intelligence caused the reduction in force. That is the first operative deadline under Connecticut’s new AI employment law, Public Act 26-15 (enacted as Substitute Senate Bill 5), and it lands in two days for any company already planning a fall layoff.
Two more deadlines follow in 2027. Frontier AI developers must stand up whistleblower reporting channels by January 1, and a full automated-decision-tool notice regime for employers arrives October 1, a full year after the WARN rule takes hold. This guide covers what changes now, what to document before the October 1 deadline, and what comes next so nobody gets caught out twice.
This is a compliance and operations guide, not legal advice. Connecticut’s Labor Department has not published detailed guidance on the disclosure form, and the specifics below come from the enacted statute and from employment-law commentary attributed by name throughout. Confirm any policy change with employment counsel before you act.
What Is Connecticut Public Act 26-15 and Who Has to Comply?
Public Act 26-15, enacted as Substitute Senate Bill 5, is Connecticut’s new artificial intelligence law covering employment, frontier AI safety and several unrelated consumer-protection provisions. Its employment sections apply to any employer already required to file a federal WARN Act notice, plus any employer or vendor that deploys automated tools in hiring, discipline or termination decisions.
The law does not create a brand-new layoff-notice regime. It attaches a disclosure duty to the WARN filing employers already make, then layers on separate obligations for AI developers and for employers using automated decision technology on a longer runway. Read the full text of Public Act 26-15 alongside the state’s existing filing process on the Connecticut Department of Labor’s WARN page.
Employers with no Connecticut-specific employment counsel are the ones most likely to miss this. Nothing in the federal WARN Act mentions artificial intelligence, so a national HR team following only the federal template will file a technically incomplete notice starting this week.
What Must a WARN Notice Disclose When AI Causes a Layoff in Connecticut?
Under Section 26 of Public Act 26-15, an employer serving a federal WARN Act notice on the Connecticut Department of Labor must also disclose whether the layoffs “are related to the employer’s use of artificial intelligence or another technological change,” in a form the Labor Commissioner prescribes. The requirement is a disclosure duty, not a ban or a delay trigger.
The statute’s own language is short, and Connecticut has not yet published the specific form referenced in Section 26. Until it does, the safest reading, per Ogletree Deakins’ analysis of the law, is to treat the disclosure as a yes-or-no attestation attached to the existing notice rather than a new standalone filing.
That single answer still has teeth. It creates a discoverable, state-held record of whether an employer characterized a layoff as AI-driven, which matters if a former employee later challenges it as pretextual or discriminatory.
How Do Employers Decide Whether a Layoff Is “AI-Caused”?
Public Act 26-15 does not define how an employer determines that a layoff relates to AI or specify a threshold of AI involvement that triggers the disclosure. That gap is an interpretation problem for employers right now, not a settled legal test, and Connecticut’s Labor Department has not issued guidance filling it as of this writing.
In practice, most RIFs involve AI at some level, whether that is a forecasting model flagging a shrinking product line, an automation rollout that eliminates a role’s tasks, or a generative-AI tool that absorbed work previously done by a team. The statute does not distinguish “AI replaced this job” from “AI made this team more efficient, so we need fewer people.”
Employment counsel quoted in Fisher Phillips’ employer to-do list recommends employers document their reasoning for whichever answer they give, since the honesty and consistency of that characterization, not just the answer itself, is what a regulator or plaintiff’s attorney will test later.
What Documentation Should HR and L&D Teams Build Before October 1, 2026?
Before the deadline, HR and compliance teams need a written record connecting each eliminated role to the business reason behind it, including any automated tool, model or analytics output that fed into the decision. That record should exist in the RIF planning file itself, not be reconstructed afterward to justify the WARN filing.
Three things belong in that file: a list of every system, human-authored or AI-assisted, that fed the layoff decision; a plain-language note on how much weight each input carried relative to ordinary business judgment; and the name of the person who made the final call and signed off on the WARN characterization.
This is the same discipline that audit evidence for compliance training programs already requires for certification records. If your organization runs training data management practices that track enrollment, attendance and completion, extending that system of record to RIF decision-attribution is a smaller lift than starting from nothing.
Build the Decision Trail Before the Layoff Meeting
Capture which specific tool, model version or analytics output touched each role slated for elimination inside the RIF planning document itself, timestamped as the decision is made, not in a memo drafted afterward to justify the WARN filing.
What Are the Three Key Compliance Dates Under Public Act 26-15?
Public Act 26-15’s employment provisions roll out on three dates over roughly twelve months: October 1, 2026 for WARN disclosure, January 1, 2027 for frontier-developer whistleblower channels, and October 1, 2027 for automated decision tool notices. Each date targets a different audience.
| Operative date | What the law requires | Who it applies to |
|---|---|---|
| October 1, 2026 | Employers filing a WARN Act layoff notice must disclose to the Connecticut Department of Labor whether the reduction in force relates to the employer’s use of AI or other technological change | Any Connecticut employer already covered by the federal WARN Act’s notice requirement |
| January 1, 2027 | Large frontier AI developers must establish and maintain an internal, anonymous reporting process for employees to flag catastrophic-risk safety concerns, with clear notice of that right given to covered employees | Developers of the largest frontier AI models, not typical employers |
| October 1, 2027 | Employers and vendors deploying automated employment-related decision technology must give employees and applicants written pre-decision notice covering the tool’s purpose, trade name, data categories and contact information | Any employer or deployer using an automated decision tool that is a substantial factor in hiring, discipline or termination decisions |
Notice the gap between the first and third dates. Connecticut gave employers a year of runway between the WARN disclosure duty and the harder automated-decision-tool regime, unusual generosity by state AI-law standards, and worth using.
How Does This Differ From the Federal WARN Act Baseline?
The federal WARN Act requires 60 days’ advance written notice of a mass layoff or plant closing from employers with 100 or more employees, filed with the state and affected workers, but it says nothing about artificial intelligence. Connecticut’s Public Act 26-15 does not change the 60-day timeline or the employee-count threshold; it only adds a disclosure question onto the notice that already has to go out.
That makes the compliance burden additive, not a parallel system. An employer with a working federal WARN process, per the federal WARN Act’s notice requirements, mainly needs one more field in that process and someone who can defend the answer given.
The risk is treating this as a check-box formality. Because the AI-relatedness answer becomes a state record, an inconsistent answer is a bigger liability than a late filing would be.
What Happens on January 1, 2027 Under the Whistleblower Provision?
From January 1, 2027, “large frontier developers,” a narrow category of companies building the most capable AI models, must establish an internal process letting employees anonymously report safety concerns involving catastrophic risk, without fear of discharge or discipline for doing so. This provision targets AI labs, not the typical Connecticut employer using AI tools in HR.
Section 2 bars a frontier developer from adopting any “rule, regulation, policy or contract” that lets it discharge, discipline or penalize an employee for reporting catastrophic-risk concerns, and requires the anonymous reporting channel by January 1, 2027. Enforcement runs through the Connecticut Attorney General, with civil penalties up to $1,000 per violation.
For most readers, this section is background rather than a direct obligation. It matters because any AI vendor you buy tools from may itself be a covered frontier developer, and its compliance posture is a fair question for your next procurement review.
What Automated Decision Tool Notice Duties Start on October 1, 2027?
Beginning October 1, 2027, an employer or vendor deploying an automated employment-related decision technology, defined as any technology that processes personal data and uses computation to produce output that is a substantial factor in an employment decision, must disclose that fact to the employee or applicant interacting with it. A separate, more detailed pre-decision notice is also required.
That pre-decision notice must cover the tool’s purpose, its trade name, the categories of personal data it analyzes and how that data gets assessed, the data’s sources, and contact information for the deployer. No notice is required where it would already be obvious to a reasonable person that they are interacting with automated technology.
The law carves out ordinary business software, including word processors, spreadsheets, spell-checkers, navigation tools, antivirus software and spam filters, plus nonmaterial job changes, scheduling and productivity monitoring. A completion-tracking dashboard is unlikely to qualify on its own; a scoring model that ranks candidates or flags employees for discipline sits squarely inside the definition.
How Does Connecticut’s Law Compare With Colorado and California’s AI Employment Rules?
Connecticut is one of three states moving on AI employment rules in the same window: Colorado’s amended AI Act takes effect January 1, 2027 with Attorney General rules still being finalized, and California’s SB 947 and AB 1883 remain pending as of this writing. Each state picked a different regulatory lever.
| Feature | Connecticut (Public Act 26-15) | Colorado (amended AI Act) | California (SB 947, pending) |
|---|---|---|---|
| AI-caused layoff disclosure | Required on WARN notices starting October 1, 2026 | Not a direct requirement | Not part of SB 947 as introduced |
| Notice before an automated tool is used | Required starting October 1, 2027 | Required under Attorney General rules from January 1, 2027 | Human-review requirement in place of a notice duty, if enacted |
| Whistleblower channel for AI developers | Required from January 1, 2027 for large frontier developers | Not addressed | Not addressed |
| Primary enforcement | Connecticut Attorney General; no private right of action for notice violations | Colorado Attorney General, with a cure period | Labor Commissioner and a private right of action, if enacted |
A multi-state employer ends up building one documentation habit and applying it everywhere. Connecticut’s WARN disclosure, Colorado’s AI Act rulemaking and California’s SB 947 bill text converge on the same ask: show your work on how an automated system influenced a decision about a real person’s job.
What Penalties Apply if an Employer Misses the Disclosure Requirement?
Connecticut treats a violation of the automated decision tool notice provisions as an unfair or deceptive trade practice, enforced exclusively by the state Attorney General with no private right of action, and it allows a 60-day cure period for violations that occur before December 31, 2027. Separate whistleblower violations by frontier developers carry civil penalties of up to $1,000 per violation.
The statute is explicit on one more point: using an automated decision tool is not, by itself, a defense against a discrimination complaint. Evidence of bias testing may count as a mitigating factor, but it does not excuse the underlying outcome.
Connecticut’s choice to route enforcement solely through the Attorney General, with no private right of action for notice violations, is narrower than California’s proposed approach, which would let affected workers sue directly. That difference matters if your organization also operates in California.
Use the Cure Period as a Compliance Checkpoint
Set an internal deadline well before December 31, 2027 to test your WARN disclosure and pre-decision notice language against a mock Attorney General inquiry, since that date is when Connecticut’s 60-day cure window for first-time notice violations closes.
What Should HR, L&D and Compliance Teams Build Right Now?
The practical response to Public Act 26-15 is four concrete steps, starting with an inventory of every system touching a hiring, discipline or termination decision and ending with a calendar that already has the 2027 deadlines on it. None of this requires new software if your existing training and compliance systems already track records well.
Step 1: Inventory every system that touches a hiring, discipline or termination decision
List every tool, marketed as “AI” or not, that scores, ranks, flags or forecasts anything tied to a person’s job, including vendor analytics inside your LMS or HRIS.
Step 2: Assign a named owner for AI-attribution documentation
Someone specific, not “HR” as a department, owns the RIF decision-attribution record and the WARN disclosure answer. Naming that owner turns a policy into a habit.
Step 3: Rewrite your WARN notice template now
Add the AI-relatedness field to your internal WARN checklist and draft the supporting rationale before a layoff is imminent, while there is time to get it right.
Step 4: Calendar the 2027 deadlines and start vendor conversations early
October 1, 2027 feels distant, but the pre-decision notice content, trade names, data categories, data sources, needs vendor cooperation that takes months to line up.
Pairing this work with AI agent governance training for the managers who will actually complete these disclosures closes the gap between having a policy and having people who can execute it correctly. A documented LMS governance framework that already assigns clear ownership over similar audit questions is the fastest place to bolt this new responsibility onto, rather than standing up a parallel process.
Is This Confirmed by Connecticut’s Own Statute or Just Law-Firm Commentary?
The operative dates, section numbers and quoted statutory language in this guide come directly from the enacted text of Public Act 26-15 published by the Connecticut General Assembly. Interpretive points, such as how to read the undefined “AI-related” standard or what a practical to-do list looks like, are attributed by name to Ogletree Deakins and Fisher Phillips, both employment-law firms publishing client alerts rather than government sources.
That distinction matters because Connecticut’s Labor Department has not yet released the specific disclosure form Section 26 references, so any description of that form is a prediction, not a confirmed fact. Treat this guide, and any coverage published before the form appears, as a reading of the statute rather than a substitute for it.
Conclusion
Connecticut’s WARN disclosure requirement takes effect in days, not months, and it is the kind of deadline a national HR team following only a federal template will miss without ever realizing it. Pull your current WARN process this week and confirm someone owns the AI-relatedness answer before the next Connecticut layoff notice goes out.
Then use the runway Connecticut gave you. A full year sits between the October 2026 disclosure duty and the October 2027 automated-decision-tool notice regime, and that time is best spent building the decision-attribution habit once so it covers both deadlines, plus whatever Colorado and California finalize in the meantime.
If your organization operates in more than one state, read this guide alongside your Colorado and California compliance tracking rather than as a Connecticut-only checklist. The documentation habit is the same everywhere; only the filing deadlines and enforcement paths differ.
FAQ
Q1. What is Connecticut Public Act 26-15?
Public Act 26-15, enacted as Substitute Senate Bill 5, is Connecticut’s new artificial intelligence law covering employment, frontier AI safety and related provisions. Its employment sections require WARN Act layoff notices to disclose AI involvement starting October 1, 2026, add whistleblower protections for frontier AI developers from January 1, 2027, and add automated decision tool notice duties for employers from October 1, 2027.
Q2. When does Connecticut's AI layoff disclosure rule take effect?
The WARN Act disclosure requirement takes effect October 1, 2026. Any Connecticut employer filing a federal WARN Act layoff notice must also tell the state Department of Labor whether the reduction in force relates to the employer’s use of artificial intelligence or another technological change.
Q3. Does Public Act 26-15 stop employers from using AI to decide layoffs?
No. The October 2026 provision only requires disclosure on the WARN notice, not permission or a delay. The law does add separate obligations, like pre-decision notice for automated employment tools starting October 1, 2027, but it does not ban AI-informed layoff decisions themselves.
Q4. How does an employer know if a layoff counts as "AI-caused" under Connecticut law?
The statute does not define a threshold, so employers must use judgment and document it. Employment-law commentary recommends recording which systems and how much weight they carried in the decision, since the honesty and consistency of the characterization matters more than which answer is given.
Q5. What automated decision technologies does Connecticut's law exclude?
The law excludes ordinary business software such as word processors, spreadsheets, spell-checkers, navigation tools, antivirus software and spam filters. It also excludes nonmaterial job changes, workplace scheduling and productivity monitoring, focusing instead on tools whose output is a substantial factor in hiring, discipline or termination decisions.
Q6. What penalties apply for violating Connecticut's AI employment law?
Automated decision tool notice violations are treated as an unfair trade practice, enforced solely by the Connecticut Attorney General with no private right of action, and a 60-day cure period applies through December 31, 2027. Whistleblower violations by frontier AI developers carry civil penalties of up to $1,000 per violation.
Q7. How does Connecticut's AI employment law compare to Colorado and California?
Connecticut is first to require AI-caused layoff disclosure, effective October 1, 2026. Colorado’s amended AI Act adds a human-review right from January 1, 2027 under Attorney General rules, and California’s SB 947 would add a human-review requirement instead of a notice duty if it is enacted. All three converge on documenting how automation influenced a decision about someone’s job.