· AI
Connecticut's New AI Layoff Disclosure Rule Shows What Cheap Regulation Looks Like
Connecticut just ran a real test of light-touch AI regulation, and the cheap part of it deserves more credit than skeptics are giving it.
As of October 1, any Connecticut employer filing a WARN Act notice for a mass layoff now has to answer one more question on the form: was artificial intelligence, or some other technological change, a contributing factor. Connecticut is the first state to attach an AI question to layoff paperwork, under the Connecticut Artificial Intelligence Responsibility and Transparency Act, known as SB 5. My instinct on new AI rules is usually to look for the hidden cost, because that is where most regulation actually bites. This one mostly does not have one, and that is worth saying plainly before I get to the part that does.
Here is the steelman for opposing it anyway. Gov. Ned Lamont blocked a broader version of this bill in each of the two prior years, siding with his own economic development office, which argued that AI rules could push companies to incorporate or expand elsewhere rather than in Connecticut. Republican state senators made the same case: regulate something changing this fast and you are legislating against a moving target, and the businesses that pay the price are the ones deciding where to locate next. Rep. Bill Buckbee called the eventual bill "theater," given how fast the technology changes under it. That is a fair worry in general. States that regulate first and iterate slowly tend to freeze in last year's technology while the industry moves on, and the compliance paperwork outlives its own relevance.
But look at what actually passed. The WARN disclosure is a single checkbox added to a form employers already have to file. It does not require an algorithmic audit, a bias study, or a new compliance department. It does not create a private right of action, so it does not open the door to the kind of plaintiff's-bar litigation industry that similar-sounding privacy laws have spawned in other states. Enforcement sits solely with the Attorney General, who has to offer a 60-day cure period before bringing a case at all. SB 5 passed the House 131 to 17 and the Senate 32 to 4, a margin that only happens when a bill has been cut down to the part almost nobody can object to. Lamont signed it after vetoing the pricier version twice. That is not an accident. It is what AI regulation looks like when the legislature actually prices in the cost of compliance instead of assuming it is free because the target is a tech company.
The honest case for this specific rule is stronger than my usual prior on regulation. Yale Law's Worker and Immigrant Rights Advocacy Clinic, representing the Connecticut AFL-CIO, testified in support, and the underlying point is reasonable: right now nobody outside a company knows how much of the AI-driven layoff narrative is real versus a convenient line in a press release. A cheap disclosure requirement generates that data without telling anyone how to build or deploy a model. If the free-market case against regulation rests on "let the market find out what works," a disclosure rule that just makes facts visible is close to the least distortive way to regulate anything. I will say it plainly: the free-market case against this particular provision is weak.
Where I come back around to worry is the follow-on rule stacked behind it for 2027. Starting October 1 of that year, any employer using automated tools that meaningfully affect hiring, promotion, or discipline decisions has to give applicants and employees written notice: the tool's trade name, what it is used for, and what personal data it analyzes. On paper this is still disclosure, not an audit mandate or a ban. In practice, figuring out which of your HR vendor's features count as decision-influencing technology, documenting what data each one touches, and writing a notice a layperson can understand is real legal and operational work. The statute itself does not define what counts as an AI-related layoff or decision, leaving the details to the Labor Commissioner, so every employer is guessing at the standard until that guidance lands.
That guesswork is cheap to absorb for a large company with an HR compliance team and an AI vendor that ships a Connecticut notice template as a feature. It is not cheap for a ten-person company that bought an off-the-shelf applicant tracking tool and has no idea what the ranking algorithm inside it actually scores on. The large vendors will turn compliance into a selling point; the small ones will leave their customers to figure it out alone. That is the regulatory-moat pattern worth watching here, not the checkbox that started it.
If you are a business owner trying to tell which category your own AI tools fall into before 2027 arrives, that question, what a given tool actually uses and whether it counts as a covered decision, is exactly the kind of call MojoAI works through with clients. Figure out which vendors already carry the compliance weight and which ones leave it on you, and you will know a lot sooner than the Labor Commissioner tells you.
Sources
References used in this article. Links also appear alongside the relevant claims.
Let's make it happen.
You don't need a polished brief. A couple of lines about where you want your company to go is plenty, and we'll come back with what we'd tackle first.
We scope the work around your goals and timeline before quoting anything, so you know exactly what you're getting.