HomeAIAI Policy & SafetyConnecticut's CART Act: The AI Rules That Took Effect on 1 October

Connecticut’s CART Act: The AI Rules That Took Effect on 1 October

Connecticut began enforcing the first parts of its artificial intelligence law on 1 October 2026, including protections for whistleblowers at large AI developers and a duty on employers to say when mass layoffs are linked to AI. The law, Public Act 26-15, began as Substitute Senate Bill 5 and runs to 39 sections. Its sponsors call it the Connecticut Artificial Intelligence Responsibility and Transparency (CART) Act.

Governor Ned Lamont signed it on 27 May and announced it on 2 June. “We can no longer wait for Washington, D.C. to do the right thing,” he said in his office’s release. The release says the law was written by the governor, Attorney General William Tong and State Senator James Maroney.

The rules already in force

Several sections took effect on 1 October, according to the Act’s text.

Frontier AI developers. The Act defines a frontier developer as one training a model with more than 10 to the power of 26 operations, and a large frontier developer as one with more than $500m in annual revenue. It protects employees who report “catastrophic risk”, defined to include risks that a model contributes to the death or serious injury of more than 50 people, or more than $1bn in damage. Large developers must set up an anonymous internal reporting channel by 1 January 2027. Breaches carry a civil penalty of up to $1,000 each, enforced by the Attorney General.

AI and layoffs. Employers filing federal WARN Act notices of mass layoffs must now also tell the state Labor Department whether the layoffs relate to “artificial intelligence or another technological change”.

AI in hiring. Using an automated employment decision tool “shall not be a defense” to a discrimination complaint, from 1 October 2026. Separate notice and information duties for these tools apply to systems deployed on or after 1 October 2027.

Content provenance. Generative AI providers with more than one million monthly users must, “to the extent commercially and technically reasonable”, embed provenance data in the images, audio and video their systems create. The Act points to the Coalition for Content Provenance and Authenticity standard as one method.

The Act also sets rules for AI subscriptions from 1 October. A companion privacy law, Public Act 26-64, has October provisions on data brokers, a ban on selling residents’ genetic data and limits on facial recognition and surveillance pricing, CT Mirror reported.

What comes next

From 1 January 2027, AI companion operators must have protocols to detect signs of suicidal ideation or self-harm and respond with resources. From 1 January 2028, platforms must get parental consent before giving minors algorithmic feeds. They must also set a one-hour daily default limit and, per the governor’s release, stop notifications to minors between 9pm and 8am.

“This is a start. This is not a finish; this is not a ceiling. This is the floor,” Maroney said, according to CT Mirror. A National Law Review alert by the law firm Bradley Arant Boult Cummings gives a practitioner’s summary.

What compliance teams have to build now

Most of the 1 October duties are about recording and reporting rather than banning anything. That makes them a data problem for the companies affected. To answer the layoff question, an employer has to be able to say why roles were cut, which means linking workforce decisions to the technology projects behind them. South Korea is considering a similar rule, as we reported in our story on its AI layoff reporting bill.

The hiring clause matters because it shifts risk onto the employer straight away. A company cannot argue that a vendor’s model made the decision. Employers using such tools will want records of how they were tested, even before the 2027 notice duties start.

The provenance rule depends on industry standards, and the qualifier “commercially and technically reasonable” will decide how strictly it bites. The thresholds, 10 to the power of 26 operations and $500m in revenue, cover only the largest model developers.

Connecticut’s approach sits between the federal and European models we compared in our piece on two playbooks for governing AI. The EU, by contrast, is building its rules through formal standards, such as the first harmonised standard under the AI Act.

Sources

Follow @tsnmediahub on X for breaking AI news as it happens.

Follow on X

Share this story

Latest stories

  1. House Members Release an Outline of the SAFE AI Act
    AI Policy & Safety

More in this category

Latest stories

  1. House Members Release an Outline of the SAFE AI Act
    AI Policy & Safety

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.