Connecticut's AI Liability Law Starts Oct. 1, 2026

Connecticut employers lose a ready defense on Oct. 1, 2026. From that date, SB 5 bars them from blaming the artificial intelligence tools they use for hiring, promotion, discipline or termination decisions when facing a discrimination claim.

Two further changes arrive the same day. The Connecticut Commission on Human Rights and Opportunities or a court may weigh evidence that an employer tested its AI tools for bias. And employers filing federal WARN Act notices with the Connecticut Department of Labor must disclose whether a layoff relates to AI or another technological change.

What arrives in the law's second phase?

The obligations do not stop there, and the timeline matters. Starting Oct. 1, 2027, Connecticut employers must provide:

• Interaction notices. Plain-language disclosures when applicants or employees interact with an AI tool, unless the interaction would be obvious to a reasonable person.
• Pre-decision notices. Written notice before a covered AI tool makes, or substantially shapes, an employment decision about a person.

HR tech vendors also come under the statute. From Oct. 1, 2027, vendors based anywhere must give employers the information needed to comply, provided their tools operate in Connecticut and were sold, licensed, marketed or configured to influence employment decisions. Employers remain responsible for compliance, though a vendor may agree in contract to handle the notices. In plain terms: the duty stays with the employer, and only the paperwork can be delegated.

How do California's 2027 deadlines reshape monitoring and layoffs?

Gov. Gavin Newsom signed a package of laws on Sept. 30 that restrict how employers use AI to monitor, evaluate and lay off workers. Starting Jan. 1, 2027, the laws cover three areas:

• Emotion tracking. Employers generally cannot use AI-enabled surveillance to read or predict employee feelings, or to collect neural data.
• Bathroom surveillance. No monitoring in workplace bathrooms, beyond narrow exceptions.
• AI layoffs. When AI or automation causes a mass layoff in whole or in substantial part, the Cal/WARN notice must state "This notice is for a technology displacement" at the top, list the automated job functions and name the technology type.

From July 1, 2027, employers cannot discipline or fire workers based on an automated system alone. Reliance on a system's score, classification or recommendation requires supporting evidence, such as supervisor evaluations, personnel files or witness interviews, plus a written post-use notice.

Two further California regimes apply. Privacy regulations effective Jan. 1, 2027 require advance notice, an opt-out or a human-review appeal exception, and answers to questions about how workplace AI was used in hiring and other major decisions. Since Oct. 1, 2025, civil rights regulations hold employers with five or more employees liable when an AI tool discriminates, with records, including decision data, kept for four years.

Why is Colorado's law on hold?

Colorado chose a narrower path. Gov. Jared Polis signed SB 26-189 in May, repealing the state's original AI law before it took effect and replacing it with lighter requirements. From Jan. 1, 2027, employers using AI to materially influence major job decisions in Colorado must provide:

• Notice at the point of interaction.
• A plain-language explanation within 30 days of a negative decision, including how to request more information.
• A right to request human review and reconsideration, when commercially reasonable.
• Compliance records kept for at least three years.

Employers and AI developers can both face liability under the state's anti-discrimination law, apportioned by each party's role; a developer is liable only when its tool was used as intended or marketed. Contracts cannot shield either party from liability for its own discriminatory conduct.

Enforcement, however, is paused. A federal court order in a suit brought by xAI bars the Colorado attorney general from enforcement, including investigations, for violations occurring up to 14 days after the court rules on xAI's preliminary injunction request. xAI has 28 days after the law's rules are finalized to file. Employers should read that as a delay, not a reprieve.

Which rules already bind employers?

Illinois has enforced HB 3773 since Jan. 1, 2026. It makes two practices civil rights violations under the Illinois Human Rights Act: using AI that discriminates by protected class in recruitment, hiring, promotion, discipline or firing, including ZIP codes as a stand-in for protected class, and using AI in those decisions without telling employees. The law generally covers employers with at least one Illinois employee for 20 or more weeks. The state Department of Human Rights proposed notice rules in May, withdrew them in June and gave no new timeline — yet the notice requirement still applies.

Illinois also enforces the Artificial Intelligence Video Interview Act, in effect since 2020. Employers using AI to analyze recorded video interviews must explain how the AI works, obtain consent, share videos only with evaluators, and delete them within 30 days on request.

New York City's Local 144 has been enforced since July 5, 2023. It requires an independent bias audit of automated hiring tools within one year of use, publication of the audit summary, and notice at least 10 business days before use, with a right to request an alternative process or accommodation.

What do the pending lawsuits test?

In Mobley v. Workday, plaintiff Derek Mobley claims Workday's AI screening rejected him from more than 100 jobs because of his race, age and disability. A federal court let the case proceed on an agency theory and conditionally certified a nationwide group of applicants 40 and older. On Sept. 14, plaintiffs asked to expand the case into a class action covering Black applicants, women, applicants 40-plus and applicants with disabilities.

Kistler v. Eightfold AI, filed in January, argues the platform scored two applicants on a zero-to-five scale using information they never provided, without Fair Credit Reporting Act notices and dispute rights. A win would place some AI applicant scores in the same legal category as background checks.

A judge dismissed Harper v. Sirius XM without prejudice on Sept. 30. The applicant claimed hiring software rejected him from about 150 jobs using race proxies such as ZIP codes and schools; the court found he had not alleged enough facts, such as statistics or a comparison group. Sirius XM denied using AI in hiring, a question the court never decided.

How should employers begin the compliance review?

Seven steps follow from the statutes now in force.

1. Inventory every AI tool that evaluates or monitors people, including features inside applicant tracking systems, assessments, chatbots and video interview tools, and flag any that produce a score, rank or recommendation used in employment decisions.
2. Map where applicants and employees are located, since remote workers can pull a state's rules into play.
3. Demand answers from vendors on data use, scoring methods and bias testing, and update contracts; Connecticut mandates vendor cooperation from 2027, and Colorado voids liability shields for discriminatory conduct.
4. Test tools for bias and repeat the checks regularly; New York City still requires its independent audit on top of in-house testing.
5. Build one notice process with state-specific versions for Illinois, New York City, California and Colorado from Jan. 1, 2027, and Connecticut from Oct. 1, 2027.
6. Set up human review and appeal rights, and let reviewers override tools with written reasons.
7. Keep vendor answers, audits, notices and review decisions on file, four years in California and three in Colorado.

With deadlines clustered in 2026 and 2027 and courts testing vendor liability now, employers that wait for regulatory clarity will find the litigation arrived first.