Connecticut AI Responsibility and Transparency Act (SB 5 / Public Act 26-15)
Connecticut · Conn. Public Act 26-15, §§ 4 to 12 (signed May 27, 2026) · effective 2027-01-01 · AI companion rules start Jan 1, 2027. The employment sections take effect Oct 1, 2026, but their notice duties apply to technology deployed on or after Oct 1, 2027. The AG's 60-day cure option covers employment violations through Dec 31, 2027, not the companion rules.
Summary
If an AI companion would make a reasonable person believe they are talking to a human, the operator must give a clear notice: visible throughout, or at the first interaction in any 24 hours and at least every three hours (every hour for users under 18). Operators must keep and post a self-harm protocol, stop the AI from claiming to be human, and for known minors block romantic or sexual interaction and manipulative techniques. From Oct 1, 2027, employers using automated employment decision technology must give written notice before the decision.
Who must comply
Operators of AI companions used in Connecticut; employers using automated employment-related decision technology, and its developers.
What you must disclose
Clear and conspicuous notice that the user is communicating with an AI companion.
When: When a reasonable person would believe it is human: static notice visible throughout, or at the first interaction in any 24-hour period and at least every three hours (every hour for users under 18).
Other duties
- Keep a protocol to detect risk of suicide, self-harm or violence, refer users to help including 988, and post it on your website.
- Stop the AI from claiming to be human, including when asked.
- For known minors: measures against romantic or sexual interaction and manipulative techniques, plus screen-time and account tools for them and their parents.
- From Oct 1, 2027: written notice before an employment decision, and plain-language notice when employees or applicants interact with the technology, unless obvious.
- From Oct 1, 2026: employers filing a federal layoff notice must tell the Labor Department if the layoffs relate to AI.
Exemptions
- Chatbots used only for business operations, customer service, patient care, education or financial services, and not marketed to consumers as a companion.
- Limited video-game chatbots, stand-alone voice assistants and narrowly tailored educational tools.
- Narrow task-specific tools whose main function is not discussing mental health, and health-support tools that do not act human or meet social or emotional needs.
- Companies that only license a model and do not set how it reaches users.
Penalties
An unfair or deceptive trade practice: up to $5,000 per wilful violation, and up to $25,000 for violating an injunction. No private lawsuits.