Connecticut Governor Lamont signed SB 5 last week. If you run an AI product touching employment, chatbots, or social platforms, your October 2026 countdown is live.
The law, formally titled the Connecticut Artificial Intelligence Responsibility and Transparency Act, runs 39 sections across five distinct regulatory tracks. First to hit: AEDT disclosure requirements on October 1, 2027.
Employers using AI tools that make or materially influence hiring, promotion, discipline, or termination decisions must notify applicants and employees about what tool is running, what data it uses, and who to contact. Developers must give deployers the compliance info they need for those disclosures. No bias audit required.
Companion chatbot rules follow on January 1, 2027. Operators need safety protocols for suicidal ideation using “evidence-based methods,” a phrase the law doesn’t define — which means legal will spend Q3 arguing what counts.
Non-human disclosures run hourly for minors, every three hours for adults. Age-detection obligations kick in when an operator “knows or has reason to believe” a user is under 18.
The SB 5 template is the one getting signed. Last year’s broader SB 2 died under a veto threat over innovation concerns; this version cleared the Senate 32-4 and the House 131-17 by stacking targeted obligations across specific use cases rather than building one comprehensive statute. Your legal team should assume this model is coming to your state before 2027.
Tactically: the AEDT section amends Connecticut’s human rights law so that anti-bias testing evidence counts in your favor even though audits aren’t required. Run the tests. The social media provisions, including a one-hour-per-day default cap on personalized recommender systems for minors, don’t take effect until January 1, 2028.
Nathan Zakhary