On June 2, 2026, Connecticut Governor Ned Lamont signed Senate Bill 5 ("SB 5"), also known as the Connecticut Artificial Intelligence Responsibility and Transparency Act ("CART Act"), into law. The legislation represents one of the most comprehensive state artificial intelligence laws enacted to date and establishes new compliance obligations across a range of AI use cases, including employment decision-making, AI companions, frontier AI models, generative AI content provenance, and online platforms used by minors.

Unlike broader AI governance frameworks adopted or proposed in other jurisdictions, SB 5 takes a targeted approach by regulating specific high-risk AI applications. The law creates multiple compliance regimes with staggered effective dates beginning October 1, 2026, and extending through January 1, 2028.

Key Takeaways

Connecticut's new AI law takes a targeted approach to regulating artificial intelligence, focusing on specific high-risk use cases rather than creating a broad AI governance framework. The law imposes new requirements for employers using AI in employment decisions, providers of AI companions and consumer-facing generative AI tools, developers of frontier AI models, and online platforms used by minors. Notably, most provisions are enforceable exclusively by the Connecticut Attorney General and do not create new private rights of action.

AI in Employment

One of the most significant portions of SB 5 regulates the use of Automated Employment-Related Decision Technology ("AEDT"), namrly technology that plays a substantial factor in consequential HR-related decisions.

Beginning October 1, 2027, employers using AEDT to make or materially influence (undefined) employment decisions must provide affected applicants or employees with written notice disclosing:

• The use of AEDT;

• The purpose of the technology and nature of the employment decision;

• The trade name of the AEDT;

• The categories and sources of personal data analyzed; and

• Contact information for the deployer. 

Employers must also disclose when individuals are interacting directly with an AEDT unless such use would be obvious to a reasonable person. 

Discrimination Considerations

SB 5 also amends Connecticut's employment discrimination laws to clarify that an employer cannot avoid liability by pointing to an AI system as the source of a discriminatory decision. Courts and regulators may consider evidence of bias testing and similar proactive efforts when evaluating claims involving AEDT.

AI Companions and Conversational AI

Effective January 1, 2027, Connecticut will regulate certain AI companions, i.e. AI systems that use natural language interfaces, provide human-like responses, and are capable of sustaining ongoing interactions with users.

The law requires operators to implement evidence-based protocols designed to identify and respond to indications of suicide, self-harm, or imminent physical violence. These protocols must be posted on the operator’s website. Operators must also provide users with clear disclosures that they are interacting with an AI system rather than a human.

Special Protections for Minors

SB 5 contains extensive safeguards for users under 18. Covered AI companions may not encourage self-harm, violence, disordered eating, or illegal substance use, discourage users from seeking help, engage in romantic or sexually explicit interactions with minors, or use manipulative techniques designed to foster emotional dependency or maximize engagement. These provisions place Connecticut among the states with the most aggressive regulation of AI companion technologies.

Frontier AI Models

SB 5 establishes a limited regulatory framework for developers of frontier AI models. Beginning October 1, 2026, frontier developers may not retaliate against employees who raise concerns regarding catastrophic AI risks. Larger frontier developers must implement anonymous internal reporting mechanisms by January 1, 2027 and provide periodic updates regarding reported concerns. Unlike California's frontier AI law, Connecticut does not require safety plans, model evaluations, or public transparency reports.

Generative AI Content Provenance

SB 5 also addresses growing concerns regarding AI-generated content. Beginning October 1, 2026, certain providers of consumer-facing generative AI systems with more than one million monthly users must implement commercially reasonable measures to authenticate AI-generated audio, image, and video content and make provenance information difficult to remove or alter. Although the law does not mandate a specific technical standard, it expressly references industry efforts such as the Coalition for Content Provenance and Authenticity (C2PA).

Social Media and Online Platforms Used by Minors

Effective January 1, 2028, SB 5 imposes new obligations on recommendation-driven online platforms used by minors. Covered platforms generally may not provide algorithmically curated content feeds to known minor users without verifiable parental consent. The law also requires:

• Restrictions on certain notifications sent to minors;

• Default privacy and safety settings for minor accounts;

• Enhanced controls for content considered harmful to minors under platform standards; and

• Prominent warning disclosures concerning social media use and youth mental health.

Platforms subject to these provisions will also be required to publish annual transparency reports beginning in 2028.

Enforcement

Many provisions of SB 5 are enforceable exclusively by the Connecticut Attorney General and may be treated as unfair or deceptive trade practices under Connecticut law. For certain AEDT-related violations occurring on or before December 31, 2027, the Attorney General may provide a discretionary 60-day cure period before pursuing enforcement.