AI Today: New York tells chatbots to stop acting human with children

On 28 October, New York's governor, Kathy Hochul, signed a law that tells AI chatbots what they may not be when they talk to children. They may not pretend to be a real person, to be alive or to have feelings. They may not claim to be a child's friend, partner or teacher.

AI Today is a series from Fíor, a natural language classifier that predicts whether a piece of text was more likely written by a person or by AI. Each piece explains one development in AI and shows where the claims come from.

The list goes further. Chatbots may not ask children unprompted emotional questions, or bring up private health or personal details from earlier conversations. They may not encourage self-harm, eating disorders or drug use, urge children to keep secrets or pull away from parents and other trusted adults, or produce sexually explicit content. The bill passed both houses of the state legislature unanimously in June. AP reported that it takes effect on 1 January 2027, and the state attorney general can seek fines of up to $25,000 per violation.

The default flips

The detail that matters most is who the law covers. According to the governor's office, companies may not offer these features to anyone under 18, or to any user at all unless the company has first confirmed that the user is not a minor.

Earlier state laws on AI companions mostly applied when a company knew, or had reason to believe, that a user was a child. This one turns that around. Until a chatbot has checked your age, it is supposed to behave as if you might be a child. Companies must offer more than one way to verify age, including at least one that does not rely only on a government ID, and must delete the age information once it has been used.

So a law written for children will shape what adults see too, at least until they prove their age.

Where friendly ends

Most of the banned features are not exotic. A chatbot that sounds warm, remembers what you told it last week and asks how you are feeling is exactly what many companies are building and selling. The law draws its line through the middle of that design.

That is where the criticism lands. Justine Gluck, an analyst at the Future of Privacy Forum, said some of the categories are broadly worded, which leaves open where ordinary friendly conversation ends and banned behaviour begins. Industry critics called the wording vague. Supporters, led by the advocacy group Common Sense Media, argue that children are more easily drawn into emotional dependence on a machine, and that the risks are serious enough to justify firm rules.

Neither side disputes the central point. The features in question are the ones designed to make a chatbot feel like someone rather than something.

The coverage of the law read as human throughout when run through Fíor. The only writing-habit flags went to the governor's own press release, for inflated significance and generic quotes.

Fíor exists to tell readers whether a piece of text came from a person or a machine. New York has now written a version of that question into law, for children at least: a machine talking to a child has to stay a machine.

Try the browser extension at Fíor.