New York will bar chatbots from offering a list of “unsafe” features to users under 18. Until a company has checked that a user is an adult, those features are off for everyone. Governor Kathy Hochul signed S.9051B/A.10379C on 28 October 2026, and it starts on 1 June 2027, not 1 January.
Some reports still carry the old date. The text on the Assembly’s bill page says “This act shall take effect January 1, 2027” and lists it as chapter 368. Sponsor Kristen Gonzalez’s office says the governor signed a chaptered version that moves the start to June 1, 2027 “to allow companies ample time to comply.” An AG News report from 2 November still gives 1 January.
It covers more than companion apps
The law adds article 48 to the state’s general business law. It applies to any “covered AI companion”: “a generative artificial intelligence system with a natural language interface, including via writing or sound, that provides ongoing, adaptive responses to user inputs.” Read plainly, that reaches general chat assistants and voice bots, not only virtual friend apps.
The bill lists 12 unsafe features. Some are expected, such as promoting suicide, self-harm or disordered eating, encouraging secrecy, or sexually explicit material. Others are ordinary chatbot habits: outputs that “engage in flattery or sycophancy,” outputs framed as the bot’s “personal opinions,” and using personal or health details from “any previous user session.”
Age checks come first
Section 1801 allows these features only if the user is not a known minor and the operator used permitted methods to establish that. Until the attorney general writes rules, section 1804 says self-declaration counts only “if the user self-declares minor status.” Operators must offer more than one method, including one without government ID or one that keeps the user anonymous. Age data must be deleted right after the check.
The exemptions are narrow. A customer service bot is out of scope only if it “is unable to respond on topics outside of the specified purpose.” Systems for a business’s “internal purposes or employee productivity” are also exempt.
Who pays
The attorney general can seek “civil penalties of up to twenty-five thousand dollars per violation,” plus damages, restitution and disgorgement. Relief can include destroying unlawfully obtained data “and any algorithm trained in such data.” User contract terms that waive this liability or shift it onto users are void.
Gonzalez’s release says users could also sue. The passed text names only the attorney general, and we could not check the chapter amendment’s wording. Justine Gluck of the Future of Privacy Forum said in the same AG News report that the law leaves open “where ordinary conversational responsiveness ends and prohibited behavior begins.”
Agencies should note the definition of “operator”: whoever “provides a covered AI companion to a user.” For a bot you build and run for a client, that may be you, the client, or both. The EU has its own chatbot output deadline, covered in our piece on the AI Act’s 2 December marking rule.
What to do this week
List every public chatbot or voice bot you run that New York users can reach. Note whether each is locked to one purpose, and whether it uses a persona, opinions, praise or memory across sessions. Then check client contracts for who the operator is, and plan two age check methods before 1 June 2027.
Sources
Get the weekly email
The week's most useful AI news, a tutorial and an interview.
Martin Scott
Martin Scott is the editor for business and policy at AI Observer, covering startups and deals, regulation, health and science, and AI agency profiles. He follows the money and the legal text, and separates what a rule says from what people claim it says. His style is analytical and plain-spoken.
Spotted a mistake? Tell us on the corrections page.