Connecticut is the newest state to join the AI regulatory arena with the Connecticut Artificial Intelligence Accountability and Transparency Act (SB 5). Parts of this law will take effect on October 1, 2026, and other provisions will be phased in until 2028. In addition to regulating AI tools in the employment field, this law also seeks to regulate other aspects of how companies develop, sell, and use AI in Connecticut.
Use of AI tools related to employment
Connecticut, along with other jurisdictions such as New York City and Illinois, is working to regulate artificial intelligence in the employment sector. The employment-related aspects of this law apply to employers who use “automated employment-related decision-making technology.” This involves using automated systems to produce output that has a significant impact on employment decisions, such as hiring and firing. Training and employment renewal decisions will also be addressed. and an output containing predictions and recommendations. It also includes classification and ranking of individuals.
Starting October 1, 2027, eligible employers must take two steps:
- Disclose information. Companies must notify applicants or employees before using automated tools to make or significantly influence employment decisions. The notification must explain in clear terms that the technology is being used and its purpose. The disclosure should also describe the nature of the decision and the trade name of the technology. It must also list the categories and sources of personal data being analyzed, as well as contact information for the implementer. The disclosure must also explain how the individual will be evaluated. Additionally, companies must disclose that an employee or applicant is interacting with a covered AI tool, unless it is obvious to someone that they are interacting with the AI.
- Avoid prejudice. SB 5 is tied to Connecticut’s discrimination law. When allegations of bias arise involving AI tools, regulators and courts will look into how companies tested their tools for unfair results and what they did if they found problems.
Violations of these automated employment requirements constitute unfair or deceptive trade practices that can only be enforced by the Attorney General, and this law does not create a private right of action. For violations of this section that occur before December 31, 2027, the Attorney General must first issue a Notice of Remedy giving the person 60 days to correct the problem before initiating litigation.
Extensive AI regulation
Despite the White House’s executive order, Connecticut is stepping in where Colorado is going. SB 5 imposes extensive obligations on those who develop or deploy AI tools, some of which mirror obligations in other states, as described below.
- AI companion. Like many other states, the law regulates AI companions. Starting January 1, 2027, operators of AI companions (which essentially means AI that mimics human-like conversations across multiple interactions) must include protocols to detect signs of suicide, self-harm, or impending violence and refer users to resources such as the 988 National Suicide Prevention Lifeline. Operators must also clearly and conspicuously communicate to users that they are communicating with an AI rather than a human. This can be done by notifying at the beginning of the interaction or throughout the interaction. Notifications must appear once every hour for users under the age of 18 (once every three hours for adult users) during a continuous session. There are additional restrictions for use of AI Companions by persons under 18 years of age. This includes ensuring your companion takes steps to prevent self-harm.
- AI content. Starting October 1, 2026, developers of AI systems that can produce synthetic digital content, such as images, audio, and video, will be required to ensure that their output is marked as AI-generated and detectable before a consumer encounters it for the first time. Labeling solutions must be technically effective, interoperable, and consistent with nationally or internationally recognized technical standards. If AI-generated audio, image, or video content forms part of an artistic, creative, satirical, or fictional work, disclosure must be limited to a form that does not interfere with the display or enjoyment of that work.
- AI subscription. Mirroring many states’ automatic renewal laws, effective October 1, 2026, companies that sell access to AI products through paid subscriptions will be required to provide consumers with a written notice disclosing their key terms and conditions, including any restrictions imposed on the consumer’s use, before the subscription begins or renews. Failure to do so is considered an unfair or deceptive trade practice under Connecticut law.
- safe harbor. The law provides for the creation of a safe harbor program operated by (up to five) independent organizations. The pilot is scheduled to begin by July 1, 2027, and companies that comply with the provisions of these organizations’ programs may be able to use their participation as evidence in the defense of private lawsuits brought alleging harm caused by AI models. Participation in government enforcement actions is not permitted. The pilot will run until June 30, 2030.
By December 31, 2026, the state of Connecticut will establish the “Connecticut AI Academy,” which will offer online courses on AI and responsible use of AI. The state also plans to take other steps, including programs to use AI to improve health outcomes and research into how AI will impact the workforce. We also explore ways to test innovative AI products under relaxed regulatory requirements.
Practice: This new law is a reminder of the state’s continued focus on AI tools. Connecticut’s focus mirrors what we’re seeing in other states, including automated tools that influence hiring and employee decisions, AI companions, and disclosures when someone is interacting with AI.
