Connecticut workplace AI law 2026 sets two-phase employer duties

Connecticut has enacted SB 5, a workplace artificial intelligence law that takes effect in two phases, according to HRMorning. The first phase begins Oct. 1, 2026, and the second follows a year later on Oct. 1, 2027.

Starting Oct. 1, 2026, Connecticut employers facing a discrimination claim cannot blame the AI tools they use to make hiring, promotion, discipline or termination decisions. The Connecticut Commission on Human Rights and Opportunities (CHRO) or a court may consider evidence that an employer tested its AI tools for bias. Employers filing federal Worker Adjustment and Retraining Notification (WARN) Act notices with the Connecticut Department of Labor must also disclose whether a layoff is related to AI or another technological change.

How the 2027 notice requirements work

The second phase, effective Oct. 1, 2027, creates two distinct notice obligations for Connecticut employers. First, employers must provide plain-language interaction notices to applicants and employees when they are interacting with an AI tool, unless the AI interaction would be obvious to a reasonable person. Second, employers must provide written pre-decision notices to applicants and employees before a covered AI tool makes, or is a substantial factor in, an employment decision about them.

The law also reaches HR technology vendors wherever they are based. Vendors whose tools are in use in Connecticut and were sold, licensed, marketed or set up to influence employment decisions must give employers the information needed to comply. Employers remain responsible for meeting the law’s requirements, though a vendor can agree in its contract to handle the notices.

What this means for HR and payroll teams

Connecticut employers should treat Oct. 1, 2026 as the first compliance milestone, not a distant date. The defense prohibition means an employer cannot argue that a biased outcome was the vendor’s fault. That shifts practical responsibility onto internal teams to document how AI tools are selected, tested and monitored before they are used in employment decisions.

Payroll and HRIS teams should also prepare for the WARN disclosure change. When a Connecticut employer files a federal WARN notice with the state Department of Labor, the filing must state whether the layoff is related to AI or another technological change. That means HR and finance leads need a process to answer that question accurately before a WARN notice is submitted.

For the 2027 notice requirements, employers should begin mapping every point where an applicant or employee interacts with an AI tool. That includes chatbots in recruiting, resume scoring systems, video interview analysis and automated scheduling tools. The “obvious to a reasonable person” exception is not a safe harbor to assume; employers should document why any skipped interaction notice qualifies.

How Connecticut compares with other state laws

Connecticut is one of several jurisdictions now regulating workplace AI. California’s law, signed Sept. 30 by Gov. Gavin Newsom, takes effect in stages starting Jan. 1, 2027. It generally prohibits employers from using AI-enabled workplace surveillance to read or predict how employees feel, or to collect their neural data. It also bans monitoring employees in workplace bathrooms, with narrow exceptions. California’s Cal/WARN notice must say at the top: “This notice is for a technology displacement,” list the job functions being automated, and name the type of technology involved when a mass layoff is caused in whole or in substantial part by AI or automation.

Colorado repealed its original AI law before it took effect and replaced it with fewer requirements. Effective Jan. 1, 2027, Colorado employers that use AI tools to materially influence hiring and other major job decisions about applicants and employees in Colorado must provide a clear disclosure at the point of interaction.

California also has civil rights regulations in effect since Oct. 1, 2025. Employers with five or more employees are responsible when an AI tool discriminates against applicants or employees, and they must keep records, including the tool’s decision data, for four years.

Key dates for multistate employers

Multistate employers should not assume that complying with one state’s law satisfies another. The definitions of covered AI tools, the notice triggers and the recordkeeping duties differ. Connecticut’s vendor information requirement is notable because it places obligations on vendors outside the state, which may affect procurement conversations with HR technology providers.

Practical steps before the deadlines

Employers should start by inventorying AI tools used in hiring, promotion, discipline and termination decisions. For each tool, document the vendor, the decision it influences and whether the vendor has committed to providing compliance information. Contracts should be reviewed to determine whether the vendor will handle notices or whether the employer must build that capability internally.

Testing for bias should begin before Oct. 1, 2026. The law allows the CHRO or a court to consider evidence of bias testing, which means employers that can show a documented testing program may be in a stronger position in a discrimination claim. Testing should be repeated when tools are updated or when the applicant pool changes materially.

Finally, HR teams should coordinate with legal counsel on the interaction notice exception. The law requires a notice unless the AI interaction would be obvious to a reasonable person. That standard is fact-specific, and employers should develop written guidance on which interactions they consider obvious and why.

Sources: HRMorning

Photo: What Sci-Fi Futures Can (and Can't) Teach Us About AI Policy (47074229724) by New America, CC BY 2.0