Connecticut has an opportunity not just to promote AI innovation, but to model AI accountability. Economic competitiveness and open government are not in conflict. In fact, durable innovation depends on public trust, and trust depends on transparency.
CCFOI supports sections of SB 86 that encourage transparency for AI use among public agencies, but suggests additional amendments that would strengthen that transparency.
See full testimony below.
SB 86, AN ACT ADDRESSING INNOVATIONS IN AND THE RESPONSIBLE USE OF ARTIFICIAL INTELLIGENCE
General Law Committee Public Hearing March 4, 2026
Good afternoon Sen. Maroney, Rep. Lamar, Sen. Cicarella, Rep. Rutigliano and honorable members of the General Law Committee. My name is David DesRoches, and I am a member of the Connecticut Council on Freedom of Information, an organization founded by journalists to advocate for stronger government accountability and transparency. I was also a member of the Connecticut Artificial Intelligence Working Group, co- chaired by Sen. Maroney and charged with developing state-level guidance on how the state should regulate Artificial Intelligence while also incentivizing business development in AI.
I appreciate the intent behind this legislation, and there are strengths worth mentioning. First, it preserves the presumption that government data is public unless otherwise restricted by law. That principle aligns with the spirit of the Connecticut Freedom of Information Act and reinforces Connecticut’s longstanding commitment to open government.
Second, the requirement that the state data plan be subject to a public hearing and posted online promotes visibility into executive branch data governance. The creation of a Chief Data Officer along with data officers at each executive branch agency is also a step in the right direction.
Third, the requirement that companion chatbot operators submit annual reports to the Attorney General, and that those reports be publicly posted, creates a baseline of public disclosure.
These are meaningful and positive provisions. However, there are several areas where transparency protections are incomplete and where public trust could be strengthened.
The most significant concern involves the proposed artificial intelligence regulatory sandbox. The bill envisions a program that would allow companies to test AI products under “reduced licensure, regulatory and other legal requirements.” Yet there is no requirement that the public be informed which companies are admitted, what rules are waived, what consumer impacts occur, or what outcomes result. If AI systems are being tested in sectors like banking, insurance, or public health, the public deserves structured disclosure. Innovation should not occur in the dark.
Second, enforcement of companion chatbot requirements is vested solely in the Office of the Attorney General. While centralized enforcement can be efficient, the bill provides no mechanism for public reporting on complaints received, investigations opened, or enforcement actions taken. Without routine disclosure, accountability depends entirely on executive discretion.
Third, the reporting requirements for chatbot operators are narrowly defined. Operators may submit only the limited information specified in the statute. That language effectively caps transparency, preventing regulators from requesting broader contextual data about user safety trends, system failures, or patterns of harm.
Finally, the bill does not address transparency when the state itself deploys artificial intelligence systems. There is no requirement for public notice when agencies use AI in decision making, no mandate for algorithmic impact assessments, and no requirement that procurement contracts or bias audits be publicly posted. As AI becomes embedded in government operations, residents should know when automated systems are influencing public services.
To be clear: this bill does not weaken existing open records law. But it leaves significant transparency questions unanswered at a moment when public trust in emerging technology is fragile. I respectfully urge the Committee to consider amendments that would:
• Require public disclosure of sandbox participants, waivers granted, and annual performance reports;
• Mandate periodic public reporting by the Attorney General on enforcement activity;
• Allow regulators to collect and publish broader, anonymized safety metrics; and
• Require transparency and notice when state agencies deploy AI tools that affect residents.
Connecticut has an opportunity not just to promote AI innovation, but to model AI accountability. Economic competitiveness and open government are not in conflict. In fact, durable innovation depends on public trust, and trust depends on transparency.
Thank you for your consideration.
David DesRoches
Member, Connecticut Council on Freedom of Information