AI Frontier Institute

An Act Concerning Online Safety.

View primary source →

Connecticut Public Act 26-15 (Substitute SB 5), "An Act Concerning Online Safety," is law. It is a broad AI and online-safety package. Section 1 requires AI subscription providers to give written notice of key terms, including usage limits and their discretion to restrict functionality, and to get the consumer's written acceptance before they enter into or renew a subscription. Section 2 protects employees of "frontier developers" (developers training foundation models above 10^26 operations) who report catastrophic-risk concerns. It bars retaliation and requires large frontier developers (over $500 million in revenue) to run an anonymous internal reporting process by January 1, 2027. Civil penalties are up to $1,000 per violation. Section 3 directs a plan for an AI regulatory sandbox. Sections 4-6 regulate AI companion chatbots, effective January 1, 2027. Operators need suicide and self-harm protocols, must not claim to be human, and must disclose that the user is talking to an AI. For users they know or suspect are under 18, operators must prevent sexual content, manipulative engagement tactics and similar harms, and must provide parental screen-time and account tools. Sections 7-12 and 13-14 cover automated employment-related decision technology. Developers must give deployers information, and deployers must disclose AI interactions and give pre-decision written notice to employees and applicants, on or after October 1, 2027. Trade secrets are protected, and the Attorney General enforces, with a cure period for violations through 2027. Using such technology is not a defense to employment discrimination claims, though anti-bias testing evidence may be considered. Section 15 requires large generative AI providers (over 1 million monthly users) to embed provenance data in image, audio and video content. Section 16 sets up a Connecticut Academy of Science and Engineering liaison, and Sections 17 and 19-22, 30 create and promote a Connecticut AI Academy at Charter Oak State College. Section 18 creates an AI working group that reports by February 1, 2027. Other provisions: an AI layoff disclosure to the Labor Department (Section 26); a UConn workforce impact study (Section 29); AI education and teacher-preparation changes (Sections 23, 27, 34); an AI-related economic strategic plan (Section 28) and talent programs (Sections 24, 25, 31, 32); a Department of Consumer Protection pilot for independent AI verification organizations (Section 33); an Attorney General technology fellowship (Section 35); and public data for AI use (Section 36). State agency AI inventories, impact assessments and use limits are covered by Sections 37 and 38. Section 39, effective January 1, 2028, restricts personalized feeds for minors on social media without age checks and parental consent. It also requires a mental-health warning label, default limits (notification curfew, one hour per day, private account mode) and annual public disclosures.

← Back to the tracker