Connecticut: New AI law creates employment-related obligations

At a glance

• Connecticut has enacted the Connecticut Artificial Intelligence Responsibility and Transparency Act (CART Act), which establishes AI-related compliance requirements that take effect in phases between October 2026 and January 2028.
• Beginning October 1, 2027, employers using automated employment-related decision technology (AEDT) in hiring or other employment decisions must provide detailed notices to affected applicants and employees.
• The legislation provides that employers may not avoid liability for unlawful discrimination by attributing an employment decision to an AI system, even where a third party provides the technology.
• The law also establishes new rules involving AI companions, frontier AI models, synthetic media, AI-based subscriptions, and online platforms used by minors.
• Many provisions are enforceable by the Connecticut Attorney General, and the legislation does not generally create new private rights of action.

---

Connecticut has enacted the CART Act, which regulates several specific uses of AI, including AEDT, AI companions, frontier AI models, synthetic media, and AI-based subscriptions, and also addresses minors’ use of recommendation-based digital platforms. Signed into law on June 2, 2026, the CART Act introduces a phased implementation timeline, with various provisions taking effect between October 2026 and January 2028.

From October 1, 2027, employers that use AI systems to make or materially influence employment-related decisions must provide applicants and employees with written notice that the technology is being used. The notice must describe the purpose of the system, identify the technology used, outline the categories and sources of personal data being analyzed, and provide contact details for the organization deploying the technology. Employers must also disclose when an individual is interacting directly with an AEDT, unless readily apparent.

The legislation also provides that employers cannot avoid liability for unlawful discrimination by attributing a decision to an AI system. Regulators and courts may consider evidence of bias testing and other proactive measures when assessing claims involving AI-assisted employment decisions.

The legislation also prohibits developers of frontier AI models from retaliating against employees who raise concerns about “catastrophic AI risks,” while larger developers will be required to maintain anonymous reporting channels and provide updates regarding reported concerns.

Please see DLA Piper’s client alert, Unpacking SB5: Connecticut’s new AI law, for details on other provisions of the law.

The CART Act reflects continued state-level activity around specific AI use cases. Employers using AI in recruitment, workforce management, or other employment processes may wish to review existing practices, assess whether any technologies may fall within the definition of AEDT, and prepare for the October 2027 disclosure obligations.