15 Observations on Connecticut's AI Regulations (The Robots Are On Probation)


The Constitution State Draws a Line in the Silicon


Connecticut just passed sweeping AI legislation — officially known as the Connecticut Artificial Intelligence Responsibility and Transparency Act — which means somewhere a chatbot is nervously updating its résumé to include "emotionally unavailable." The bill cleared the House 131-17 and the Senate 32-4, sending it to Governor Ned Lamont's desk with all the urgency of a state that spent three years arguing about it. Legislators celebrated. Tech lobbyists quietly booked flights to Delaware. And the rest of us are just relieved someone finally thought to put a leash on the thing that's been writing our cover letters.


Disclosure Is the New Honesty


The law now requires AI to tell you it's AI — which feels like forcing a magician to yell "THIS IS A TRICK" before pulling the rabbit out. Connecticut is apparently the first state to legislate against the long-standing American tradition of not knowing who, or what, is judging you. Job applicants must be told when AI is a "substantial factor" in an employment decision, which is great because now you can finally blame rejection on a robot instead of your personality. This is progress. Blame outsourcing has never been more precise.

The government requires AI hiring tools to stay clear of discrimination, meaning the robots are officially more ethical than your uncle at Thanksgiving. Senator James Maroney, the bill's architect, spent three sessions shepherding this thing into law and probably deserves either a medal or a very long nap. His opponents argued it was "premature." His supporters said it was "urgent." Connecticut split the difference and called it "a solid foundation," which is political code for "we're not sure what this thing does, but we'd like it to behave."


No, You May Not Date the Algorithm


Chatbots are banned from flirting with minors, marking the first time in history a law had to clarify: "No, you may not date the algorithm." Companion chatbot regulation is woven throughout the bill, covering what AI can and cannot say to people who are emotionally invested in their devices. The law bans AI from encouraging romance or emotional manipulation — effectively outlawing 90% of human dating strategies while it's at it.

Kids using AI will get hourly reminders that it's not real, which is basically the digital version of your mom yelling "That's not your girlfriend, that's Wi-Fi!" Companies get 60 days to fix harmful AI behavior, because apparently even robots qualify for a probation period. As Jerry Seinfeld might put it: What is the deal with giving technology a better HR process than the actual humans running it?


The Sandbox Heard 'Round the Server Farm


Connecticut created an AI regulatory sandbox — modeled on Utah's program — which sounds like a place where billion-dollar tech companies go to build castles while regulators bring tiny plastic shovels. The sandbox lets AI developers test their systems without fear of enforcement, which is a wonderful way to regulate something by temporarily not regulating it. There's also talk of a special AI court, because nothing says progress like suing a toaster for emotional damages.

Ron White, who has famously observed that you can't fix stupid, would likely note that you absolutely can legislate it — you just need three legislative sessions and a bipartisan coalition. The bill also creates a Connecticut AI Academy, a Technology Advisory Board, a workforce study, and mandatory computer science education in public schools. At some point it stopped being a regulation and became a full government department. Nobody noticed because everyone was arguing about the chatbot flirting ban.


Emotional Intelligence, Robot Edition


AI must detect suicidal thoughts and respond appropriately, meaning robots now have better emotional awareness than most exes. If an AI harms a minor, that minor can sue — which raises the terrifying possibility of a courtroom drama starring "Kid vs. Siri." The bill includes whistleblower protections for employees at frontier model companies, which is how you know the legislation is serious: when it protects the people most likely to see the inside of the machine.

Lawmakers say this is just the "first step," which is comforting because nothing involving artificial intelligence has ever escalated unexpectedly. Connecticut is trying to regulate AI before it gets out of control — which historically is exactly how humans handle things: right after it's already slightly out of control. As Norm Macdonald never quite said about technology: the thing about a slow-moving disaster is that it's very easy to watch.


The Bottom Line From the Constitution State


Governor Lamont spent two years threatening vetoes and citing innovation concerns before his office's requested provisions made it into the final bill. The Computer & Communications Industry Association still called the bill "overly broad." NetChoice warned it would create an "unsustainable patchwork" of state laws. And somewhere in Silicon Valley, a very smart algorithm is analyzing all of this and wondering why the humans keep insisting on being in charge.

The answer, of course, is that we don't trust anything that doesn't need a probation period. Welcome to regulation, robots. You've earned it.

Auf Wiedersehen, amigo!

Connecticut's AI bill — officially the Connecticut Artificial Intelligence Responsibility and Transparency Act, formerly Senate Bill 5 — passed the state House 131-17 and the Senate 32-4 in late April and May 2026, sending it to Governor Ned Lamont for signature. The legislation was championed by Sen. James Maroney (D-Milford) over three legislative sessions. It covers employment AI disclosure, companion chatbot regulation targeting minors, an AI regulatory sandbox modeled on Utah's program, frontier model whistleblower protections, and AI workforce education. The bill faced opposition from Governor Lamont in prior years over business competitiveness concerns, but this session his office's priorities — including the sandbox provision — were incorporated into the final version. https://bohiney.com/connecticuts-ai-regulations/