Connecticut AI Legislation in 2026

The Connecticut Picture

Connecticut is, in my assessment, the most persistent state-level AI legislator in the country, and that persistence finally converted into law in 2026. The throughline is Senator James Maroney, who drove a comprehensive AI bill in three consecutive sessions. The 2024 SB 2 cleared the Senate and died in the House. The 2025 SB 2 cleared the Senate again and stalled when Governor Lamont signaled he preferred a multistate approach over a Connecticut-first mandate. The 2026 vehicle, SB 5, finally broke through, passing both chambers with bipartisan margins and earning the governor's signature on May 27, 2026, as Public Act 26-15. What changed was packaging. Rather than a standalone algorithmic-discrimination statute resembling Colorado's law, Maroney folded AI obligations into a broader online-safety frame covering minors, chatbots, frontier models, and provenance, which proved more politically durable.

What makes Connecticut distinctive is not just durability but Maroney's role as a multistate coordinator. He has anchored a consortium of state lawmakers seeking interoperable AI rules, which is precisely the framework Lamont said he preferred. The result in SB 5 is a law that is broad in surface area but more risk-tiered and sector-specific than a single sweeping discrimination mandate. It reaches employment decision tools, consumer-facing companions, generative-AI provenance, and frontier developers above a 10^26 FLOP training threshold, with whistleblower protections tied to catastrophic-risk reporting. Layered on top of the 2023 government-AI inventory law, Connecticut now regulates both its own public-sector AI use and significant slices of the private market.

For builders and businesses, the practical posture is staged compliance rather than a single deadline. I would treat October 1, 2026 as the baseline activation date for most provisions, January 1, 2027 for AI companion and chatbot safety duties, and October 1, 2027 for the automated employment decision obligations that apply to newly deployed systems. Employers using AI in hiring, promotion, discipline, or discharge should map where AI is a substantial factor, prepare candidate and employee notices, and secure developer compliance documentation. Developers shipping generative tools into Connecticut should plan for provenance embedding. Frontier developers above the training threshold, especially large developers above $500 million in revenue, should stand up safety and whistleblower protocols now.

Finally, do not read SB 5 as the whole picture. Connecticut handled synthetic intimate imagery and deepfake harms through a separate companion measure, HB 5312, the Take It Down Act, which adds a private right of action and Attorney General enforcement. The election-deepfake and nonconsensual-image provisions that lived inside the failed SB 2 versions migrated into this dedicated track. Companies should therefore treat Connecticut as a multi-statute environment: the 2023 act for government use, Public Act 26-15 for commercial AI and online safety, and HB 5312 for synthetic-media harms.