Connecticut’s AI Responsibility and Transparency Act, a 74-page omnibus Gov. Ned Lamont signed in May, began phasing in on October 1. Public Act 26-15 arrived the same week California closed its deadline with 11 new AI laws, extending a state-level patchwork that now routes through the Connecticut Attorney General and carries CUTPA penalties up to $5,000 per violation.

Four obligations are live. Paid AI subscriptions sold into Connecticut must deliver written terms and secure written acceptance before sign-up or renewal, per Superpower Daily. Generative-AI providers with more than one million monthly users must embed consumer-accessible, tamper-resistant provenance data, the National Law Review and ComplianceHub note, pointing to C2PA-style standards. WARN notices for layoffs must now disclose whether AI or other technological change drove the reduction, Davis Polk reports.

The one that reshapes the customer-acquisition stack is quieter. Connecticut’s anti-discrimination statute now says that use of automated employment-related decision technology “shall not be a defense” to a discrimination claim. The full AEDT notice regime doesn’t bite until October 1, 2027, but the amendment operates today. Any business running AI to score prospects, filter candidates, or route customer-facing applicants can’t point at a vendor’s model as cover.

The structural read: the provenance rule aims upstream at large providers, so a five-to-thirty-person marketing team’s compliance work is really diligence on whether its content tools ship conformant marking. The anti-discrimination change aims downstream, at deployers. Vendor risk and deployer risk have been formally decoupled.

More follows. Companion chatbot rules and frontier-developer whistleblower channels activate January 1, 2027, and the AEDT notice regime nine months after that. Connecticut is sequencing the way Colorado staged its revised ADMT and chatbot rules: small live obligations first, the heavy regime on a visible countdown.

Sources

Sources