Connecticut Artificial Intelligence Responsibility and Transparency Act (CART Act; An Act Concerning Online Safety)
CT CART Act
Artificial intelligence · Children · Employees
A broad 2026 AI law signed May 27, 2026. For privacy purposes it requires AI companion operators to detect self-harm risk and bars manipulative, romantic or sexual interactions with minors; requires employers using automated employment decision tools to tell workers and applicants, including what personal data is analyzed; makes clear that using such tools is no defense to discrimination; and from 2028 bars personalized recommendation feeds for minors without parental consent, with limits on age-verification data use.
- Where
- Connecticut
- Citation
- P.A. 26-15 (SB 5, 2026), esp. ss. 4-6 (AI companions), 7-12 (automated employment decision technology), 13-14 (Conn. Gen. Stat. 46a-60(b), 46a-81c), 15 (provenance data), 39 (recommendation feeds and minors); not yet codified
- Status
- Enacted, not yet in force
- In force since
- 2026-10-01
- Enforced by
- Connecticut Attorney General (CUTPA); Commission on Human Rights and Opportunities for employment discrimination provisions
- People can sue
- Limited
- Penalties
- Violations of the AI companion (ss. 5-6) and employment-technology (ss. 8-11) provisions are CUTPA unfair trade practices enforced solely by the Attorney General, with a discretionary notice-and-cure for employment-technology violations through December 31, 2027 (s. 12). Section 39 violations are CUTPA violations without an AG-only clause. Discrimination claims proceed under the CFEPA.
- Applies to
- Operators of AI companions offered to users in Connecticut (ss. 4-6, effective 2027-01-01)
- Developers and deployers of automated employment-related decision technology used for employees or applicants in Connecticut (ss. 7-12; deployer duties apply to deployments on or after 2027-10-01)
- Employers, under the Connecticut Fair Employment Practices Act (ss. 13-14)
- Operators of platforms that algorithmically recommend user-generated media, for users under 18 (s. 39, effective 2028-01-01)
What a privacy notice must say
- Deployers of automated employment-related decision technology must disclose to employees and applicants when they are interacting with it.P.A. 26-15, s. 9 · From 2027-10-01
- Before a consequential employment decision, give the employee or applicant written notice of the tool's use, purpose, trade name, the categories and sources of personal data it analyzes and how that data is assessed, and deployer contact information.P.A. 26-15, s. 10 · From 2027-10-01
Practices it requires
- AI companions must run evidence-based protocols to detect expressions of suicide, self-harm or violence and refer users to resources such as 988, and must not claim to be human.P.A. 26-15, s. 5(a) · From 2027-01-01
- Where the operator knows or has reason to believe a user is a minor, prevent the AI companion from encouraging self-harm, violence, disordered eating or substance use, engaging in romantic or sexual interactions, or using manipulative engagement techniques; offer parental controls.P.A. 26-15, s. 6 · From 2027-01-01
- Platforms may not serve personalized feeds of user media to minors without verifiable parental consent; information collected to determine age or obtain consent must be used only for that purpose and deleted immediately.P.A. 26-15, s. 39(b)(1), (b)(3) · From 2028-01-01
Other duties
- Use of an automated employment-related decision technology is not a defense to a discrimination complaint; evidence of anti-bias testing may be considered.P.A. 26-15, s. 13 (Conn. Gen. Stat. 46a-60(b)) · From 2026-10-01
Sources
- Official text
- P.A. 26-15 (SB 5, 2026), Connecticut General Assembly
- Attorney General press release, Sept. 16, 2026
Checked against these sources on 2026-09-25 by research agent (Claude), primary sources.
Unverified: The 'CART Act' short title is taken from the Attorney General's press release; the act's own title is 'An Act Concerning Online Safety'. Whether any section formally names the CART Act was not confirmed. | Only privacy-relevant sections were reviewed; section 1 (AI subscriptions), section 2 (frontier model catastrophic-risk duties), section 15 (generative AI provenance data) and section 38 (state agency AI use) were not summarized in detail.
Research reference, not legal advice.