Connecticut AI Law Turns Hiring Tools Into Paper Trails

The Connecticut artificial intelligence (AI, machine-based systems that can generate recommendations, classifications or scores) law signed by Ned Lamont, Connecticut’s governor, puts employers on a new clock: tell workers and job applicants when automated decision tools shape hiring, promotion, discipline, firing or other workplace outcomes, and identify the data those systems use. The deeper shift sits upstream, in vendor contracts and layoff notices.

Public Act 26-15, listed in the governor’s May 29 bill notification as signed on May 27, gives the state attorney general enforcement power and leaves workers without a new private lawsuit right for the notice provisions. Most employer notice duties apply to tools deployed on or after October 1, 2027.

The New File Starts Before a Job Decision

Connecticut does not wait for a final hiring decision to ask for transparency. The statute covers automated employment-related decision technology (AEDT, software that processes personal data and produces a prediction, recommendation, classification, ranking, score or similar output that meaningfully affects a workplace decision). That reaches the machinery behind resume screens, promotion rankings, discipline recommendations and discharge decisions when the output serves as a substantial factor.

The covered decision list goes beyond hiring. It includes promotion, renewal of employment, discipline, discharge, training or apprenticeship selection, tenure, privileges and conditions of employment. The law carves out nonmaterial changes in job tasks, hours or work assignments, plus workplace health and safety, scheduling, planning and productivity monitoring.

When a covered tool generates output used for a decision, the employer must give written notice before the employment decision. The notice requirements in the 74-page Public Act 26-15 text include:

• The purpose of the tool and the nature of the workplace decision.
• The trade name of the technology.
• The categories of personal data the system will analyze or process.
• How that data will be assessed in reaching the decision.
• The sources of the personal data and contact information for the employer.

The Connecticut wrinkle sits in the contract hook. A developer that sells or licenses a tool for employment decisions must provide the information a deployer needs for those notices, unless the tool was not marketed or configured for that purpose. The law also allows a developer to assume some notice duties through a binding contract, which makes procurement the first compliance step.

Layoff Notices Become an AI Paper Trail

The sharpest workforce provision may be the part that looks least like a hiring rule. Starting October 1, 2026, an employer that sends a covered layoff notice to the Connecticut Labor Department must also disclose whether the layoffs are related to the employer’s use of AI or another technological change.

Connecticut’s WARN filing guidance already tells employers to send written notice to the state Dislocated Worker Unit, local officials and affected workers or their representatives. The new state field adds a question that federal law did not ask: whether technology helped cause the reduction.

That field matters because it moves AI displacement from earnings calls, office rumors and severance letters into a state filing process. The statute does not promise a clean public dashboard, and employers will still fight over the word related. But the form creates a record that can be counted, compared and challenged later.

Connecticut Borrows the Patchwork and Adds a Contract Hook

Connecticut enters a crowded but uneven state and local rulebook. New York City focuses on bias audits for automated hiring and promotion tools. Colorado’s new automated decision-making technology (ADMT, a broader term for systems that help make consequential decisions) law reaches beyond employment. California has treated workplace algorithms as an anti-discrimination and records problem.

The table shows why Connecticut is more than a simple notice state. A vendor cannot market an employment tool and then leave the employer unable to name the data categories or sources the statute requires. That turns vendor questionnaires, service agreements and product documentation into legal evidence.

The Trade Secret Clause Leaves a Narrow Window

The notice right has an edge and a ceiling. Connecticut asks for data categories, sources and an explanation of how personal data will be assessed. It does not require disclosure of source code, model weights, full training sets or information protected as a trade secret under state or federal law.

Even that limit comes with a paper requirement. If information is withheld under the trade secret clause, the person withholding it must send a notice saying information is being withheld and giving the basis for the decision. That will matter when a candidate receives a thin notice and asks whether the missing pieces were protected or merely inconvenient.

The anti-discrimination piece lands earlier than the main employee notice regime. On October 1, 2026, employers cannot point to use of an AEDT as a defense to a discrimination complaint. A state human rights commission or court may consider anti-bias testing and similar proactive efforts, including their quality, recency, results and the employer’s response.

Workers still get no new private lawsuit right for the notice provisions. The attorney general’s office becomes the bottleneck for those sections, with violations treated as unfair or deceptive trade practices. For employers, the safest file will not be a glossy AI policy. It will be a dated record showing what the tool did, what data it used, who reviewed it and what changed after testing.

The State Wants More Than Compliance Checklists

Read as a workforce bill, the law reaches past employer notices. It creates or repurposes state programs around AI adoption, talent, training, research and oversight. That matters because Connecticut is not only regulating workplace technology from the outside. It is trying to build a state file on where the technology hits workers.

The Department of Economic and Community Development gets an Artificial Intelligence Policy Office and an Artificial Intelligence Learning Laboratory program. The Technology Talent and Innovation Fund Advisory Committee must partner with public and private higher education institutions and training providers on AI programs. The University of Connecticut’s Institute for Municipal and Regional Policy must study AI’s effect on the state’s workforce, including entry-level jobs and groups underrepresented in the workforce.

Another provision directs the Department of Consumer Protection to approve no more than five independent verification organizations for a pilot program. Those groups would assess specified harms and risks tied to AI models under state-approved scopes. That pilot will not solve hiring bias by itself, but it gives Connecticut a route to test whether third-party verification can carry legal weight without becoming a rubber stamp.

The Compliance Clock Has Two Start Lines

The first line arrives on October 1, 2026. That is when the WARN disclosure and the no-AI-defense discrimination language begin. Companies using automated tools in Connecticut employment decisions cannot treat the later notice trigger as permission to wait on inventories, vendor files or testing records.

The second line arrives on October 1, 2027, for the main pre-decision notices tied to new deployments. The distinction matters. Tools already in use may raise discrimination and layoff-record issues before a newly deployed system triggers the full notice packet for workers and applicants.

The law also gives the attorney general discretion, for violations of the AEDT notice sections occurring on or before December 31, 2027, to issue a cure notice when a cure is possible. If the person fails to cure within 60 days, the attorney general may sue. That soft landing is narrow, and it disappears quickly.

If enforcement starts with quiet cure notices, Connecticut’s law will first operate like a procurement standard. If the attorney general’s office moves to public cases, every human resources technology contract in the state becomes a risk document before the first score is generated.

Disclaimer: This article is for informational purposes only and does not provide legal advice. Employment, privacy and AI compliance rules carry legal risk and may vary by facts and jurisdiction. Employers, workers and vendors should consult a qualified attorney or compliance professional. Dates and statutory details are accurate as of publication.