Committee advances two AI bills seeking guardrails, training, and state data access

The General Law Committee on March 21 approved a pair of bills aimed at regulating and supporting the use of artificial intelligence.

SB 2, described by sponsors as a “protect, promote and empower” package, would create transparency requirements for higher-risk AI systems, require notice when AI is used for important decisions, provide appeal and correction rights when automated decisions rely on inaccurate data, and criminalize nonconsensual intimate-image deep fakes. The bill would also authorize outreach and education by the attorney general, create training programs for students and small businesses and promote incentives and tools to help compliance.

Representative Lamar summarized the bill’s three pillars: “protect our protections ... promote ... (by) working with CASE to develop a tool to help people comply ... create a regulatory sandbox ... and empower ... the state government to use AI to more efficiently serve our constituents.”

SB 12‑49, also advanced, would direct the state chief data officer to examine and make state datasets available for AI training under prescribed policies, create a regulatory sandbox to let businesses test models under oversight, and authorize a Connecticut Innovations investment fund focused on AI and quantum technologies. The bill also clarifies that AI is not a defense to illegal discriminatory conduct.

Committee debate and concerns

Senator Kissel, while praising Connecticut’s leadership in data policy, said he would vote no on SB 2 because he worried that early, detailed disclosure obligations might drive innovators and investment away. He argued for caution so as not to “strangle the baby in its cradle” and urged the committee to focus enforcement on suspect outcomes rather than forcing disclosure of proprietary model details.

Other members, including sponsor Senator Maroney, said the bills balance protections with innovation. Maroney noted business surveys showing many companies delay AI adoption until regulatory guardrails are clear, and argued that predictable rules could increase adoption and investment. Several members asked about protections for small users and "off-the-shelf" exemptions; sponsors said the bills include an exemption for businesses that simply use a covered model where the developer assumes liability.

Procedure and votes

SB2 (enact concerning artificial intelligence) was moved to the floor (JFS to the floor) following debate. A roll call recorded primarily affirmative votes with a few members voting no; the committee recorded individual yes/no votes during the roll-call (see Actions).

SB 12‑49 (addressing innovations in AI) was also advanced by the committee and recorded a favorable roll call.

Why it matters

Lawmakers said the bills aim to reduce discriminatory outcomes and require testing and transparency for AI used in important decisions, while also investing in workforce training and a state-led sandbox to support startups and developers. Supporters argued the measures help businesses adopt AI responsibly; opponents warned of overbroad disclosure requirements that could discourage investment.

What comes next

Both bills have been approved by the committee for referral to the floors of their houses. Sponsors said they will continue discussions with stakeholders on exemptions and enforcement detail as the bills proceed.