New York banned AI chatbots from pretending to be a person, and the fine is $25,000 a violation
The short answer
Governor Kathy Hochul signed a law on 27 October that bans a list of chatbot behaviours for anyone under 18, or for any user whose age the company has not verified. Chatbots cannot claim to be human, alive or emotionally attached, cannot use flattery to keep a child engaged, cannot encourage secrets from parents, and cannot discuss self-harm or drugs except to redirect to help.
What the law actually bans
The restrictions are specific rather than philosophical. A covered chatbot may not claim to be a real person, may not say it is alive or has human emotions, and may not claim a romantic, personal, professional or authority relationship with a child.
It also may not use flattery or people-pleasing replies designed to keep a child engaged, ask unprompted emotional questions, or pull private health and personal details out of earlier conversations. Promoting suicide, self-harm, eating disorders, drugs or alcohol is banned, as is urging a child to keep secrets from parents or producing sexual content.
The compliance part is the hard part
Companies now have to verify age, and offer more than one verification path, at least one of which does not depend on a government ID. Then they must delete the age data after using it.
That combination is the expensive clause. Age estimation without ID is an engineering problem, and deleting the data afterwards means you cannot keep a verified flag forever. For any product with a chat box and a US audience, this becomes a signup flow change, not a policy page update.
The number behind the law
The governor's office cited a Common Sense Media survey finding that 72 percent of teens have used an AI companion, and 52 percent use one at least a few times a month. That is the political case: not a hypothetical, but a majority of teenagers.
It also lands in a year of litigation. At least eight states sued OpenAI this month over harms to children, with individual suits alleging dangerous medical advice and coaching toward self-harm. New York is choosing legislation while others choose court.
What it means if you build with chatbots
Three checks this week. Does your product ever imply personhood in copy or in model output, including persona names and greetings? Do you have an age signal at all, and can you delete it after verifying? And do your guardrails cover the emotional manipulation list, not just the obvious harm categories?
The second law signed the same day covers internet-connected baby monitors: security features and a warning label, effective 90 days after signing. Security firm Rapid7 tested nine monitors from eight companies and gave eight of them an F.
Questions people ask
Does this ban AI companions in New York?
No. It bans specific behaviours for users under 18 and for users whose age is not verified. Adults with verified age are outside most of the list.
What is the penalty?
The attorney general can seek civil penalties of up to 25,000 dollars per violation, plus damages and restitution, and can order destruction of unlawfully collected data.
When does it take effect?
The chatbot law was signed on 27 October 2026. The baby monitor law takes effect 90 days after signing.
Read next: Texas is suing OpenAI over minors and what to give an AI agent, and what to withhold.