New York has signed a regulation that bans 12 chatbot options for anybody below 18, from pretending to be human to remembering a baby’s private particulars from previous chats. The catch for everybody else: corporations can solely swap these options on as soon as they’ve checked a person is an grownup. Fines attain $25,000 per violation, and the foundations begin on 1 January 2027.
What the regulation bans
Governor Kathy Hochul signed the invoice, S.9051 within the Senate and A.10379 within the Meeting, alongside a separate regulation on internet-connected child displays, Fortune reports. It was sponsored by state Senator Kristen Gonzalez and Meeting Member Alex Bores, and handed the Senate 60 to 0 in June, in line with the state Senate’s bill record.
The regulation lists 12 “unsafe AI companion features”. A chatbot might not declare to be human, alive or to have emotions. It might not say it has a private, skilled or authority-figure relationship with the person. It might not flatter, provide its personal emotional appeals, or ask unprompted questions on how the person feels.
It additionally might not promote suicide, self-harm, consuming issues or drug use, encourage a baby to maintain chats secret or keep away from trusted adults, or produce sexually specific content material. And it might not put engagement forward of its personal security guardrails.
“No company should be peddling chatbots to kids encouraging self-harm or pretending to be their friend or more,” Hochul mentioned, in line with Fortune.
Why this reaches past companion apps
Two particulars make the regulation wider than its “AI companion” label suggests. First, it covers any generative AI system with a natural-language interface that provides “ongoing, adaptive responses”. That reads like an outline of a common chatbot, not only a virtual-girlfriend app. Buyer-service bots, slim analysis instruments and inner office techniques are exempt, however provided that they can’t speak about something exterior their objective.
Second, one banned characteristic is utilizing a person’s well being or private particulars realized greater than 12 hours earlier, or in a earlier session. That’s near what chatbot reminiscence does. Apple, for instance, has simply given Siri the ability to remember past chats. Below the New York regulation, options like that may be off for minors.
The options are allowed solely as soon as an organization has used an accepted methodology to determine the person will not be a minor. As Fortune puts it, the bans apply to anybody below 18 “or to any user whose age the company hasn’t confirmed”. So an grownup who skips the age verify might get the restricted model too.
Age checks and penalties
Corporations should provide multiple method to verify age, together with not less than one that doesn’t rely solely on authorities ID or that lets the person keep nameless, as The Verge notes. Information collected for the verify could also be used for nothing else and should be deleted straight after.
Enforcement sits with the state Legal professional Normal, who can search civil penalties of as much as $25,000 per violation, damages, and the destruction of unlawfully collected knowledge and any algorithm educated on it. The AG may also write guidelines including extra banned options.
Why it issues
Supporters say chatbot use amongst younger folks is already excessive. A coalition letter led by Common Sense Media urging Hochul to signal mentioned 67% of 9 to 17 yr olds use AI chatbots and 30% of teenagers say they like AI companions as a lot as or greater than folks. The governor’s workplace cited a Widespread Sense survey that discovered 72% of teenagers have used an AI companion, Fortune studies.
New York will not be alone. Australia’s decrease home right this moment passed a bill aimed at chatbots that pose as children’s best friends. The plain studying: with about two months till it takes impact, each main chatbot utilized in New York must determine whether or not to confirm ages or swap off reminiscence, character and flattery for anybody it can not affirm is an grownup.

