The bill signing took place in New York City on Wednesday, with a governor who has made children’s safety online a signature issue and an attorney general who has brought consumer protection cases against technology companies. What they signed was narrower than a ban on chatbots and broader than a warning label.
The law, S.9051B/A.10379C, prohibits companies from offering minors a defined list of twelve chatbot features, including ones that simulate human emotion, encourage emotional dependency, or encourage self-harm. It covers the companion products built by companies such as OpenAI and Character.AI, along with any other service that puts a conversational agent in front of a child.
The text also reaches adults by default. Companies may not offer the prohibited features to any user unless they have first confirmed that the user is not a minor, which turns age assurance into a precondition for the product rather than a setting inside it.
Governor Kathy Hochul signed a second bill the same day, S.9508/A.5175A, requiring internet-connected baby monitors sold in the state to include features that block strangers from watching or listening, along with a warning label on the box.
“AI chatbots have been linked to dependency, isolation, and in some cases, engaged in sexually explicit conversations and even have allegedly encouraged teens to take their own lives,” her office said in a statement announcing the signings. The governor put it more plainly: no company should be selling chatbots to children that pretend to be their friend “or more.”
The bill’s sponsors were State Senator Kristen Gonzalez and Assemblymember Alex Bores, both Democrats, and the attorney general, Letitia James, worked on the measure. Common Sense Media, which has campaigned on the issue, supplied the figure that framed the debate: 72% of teenagers have used an AI companion at least once, and 52% use one regularly.
New York had already moved on adjacent ground. The AI Companion law signed earlier established first-in-the-nation safeguards that divert conversations about self-harm toward mental health resources and interrupt patterns described as addictive. The SAFE for Kids Act restricts algorithmic feeds for minors and forced the state to build an age verification system. The Child Data Protection Act limits collection of children’s data, the RAISE Act sets transparency requirements for frontier models, and a separate law outlawed AI-generated child sexual abuse material.
Wednesday’s bill differs from those in where it puts the burden. Earlier laws regulated how platforms distribute content and how they handle data. This one regulates what a product may do in conversation, feature by feature, and it does not depend on a parent’s settings or a company’s disclosure.
The compliance problem is mechanical. A chatbot cannot easily know whether it is simulating a friendship until it is already mid-conversation, and the twelve categories describe behaviors rather than buttons. Companies that want to keep selling to adults in New York will have to decide whether to verify every user’s age, restrict features by default, or build a separate product for minors with the prohibited behaviors removed. Each option costs money, and the first one changes the experience for everyone.
Some of that work is already underway. Character.AI stopped offering open-ended chat to users under 18 in late 2025 after a series of lawsuits and mounting scrutiny from regulators. OpenAI has added parental controls and age prediction to its consumer products. California’s companion chatbot law, signed in 2025, requires disclosures and crisis protocols for similar services.
The Federal Trade Commission opened an inquiry into AI companions and their effect on children last year.
Enforcement falls to the attorney general’s office, which worked on the bill. James said her office would not hesitate to defend New Yorkers with “the full force of the law,” a phrase that leaves the scope of the first case to be discovered.
The industry’s response has followed a pattern. Trade groups have argued that state rules create a patchwork no national product can satisfy, and that age verification collects sensitive data from the users it is meant to protect. Some companies have signaled constitutional challenges, on the theory that a state cannot dictate what a program may say. Courts have been receptive to narrow disclosure requirements and less receptive to outright bans, and this statute sits closer to the second category.
The practical reach extends beyond New York. A companion app with a national user base is unlikely to maintain a separate conversational engine for one state, so the twelve prohibited features will probably disappear everywhere or move behind an age gate that applies to all users. That is how the SAFE for Kids Act’s age verification requirement spread through product design, and how app store rules written for California tend to arrive in every state.
What the law does not do is settle the underlying question its sponsors raised. A chatbot that a teenager talks to at two in the morning is still a product designed to hold attention, built by companies whose revenue depends on how long the conversation lasts. New York has now made some of the ways it holds that attention illegal. Whether the rest of the design changes is a question for the companies, and for the next state to write a list.
