New York Just Signed the Broadest AI Chatbot Safety Law in the US — What Changes on 1 January 2027

New York has signed one of the broadest AI chatbot safety laws in the United States. Governor Kathy Hochul signed bill S9051-B / A10379-C into law on 28 October 2026. From 1 January 2027, chatbot operators in New York must not offer 12 defined "unsafe AI companion features" to users under 18 — or to any user whose adult status has not been established through age assurance. The law does not ban ChatGPT, Claude or Gemini. It bans a set of features that make a chatbot act like a human friend, boyfriend, girlfriend or confidant, and that encourage dangerous behaviour. How far it reaches in practice is already disputed.

The law adds a new Article 48 to New York's General Business Law. It was sponsored by State Senator Kristen Gonzalez and Assemblymember Alex Bores, developed with Attorney General Letitia James and Common Sense Media, and passed with large majorities: the Senate 60–0 on 4 June 2026, and the Assembly 137–0 on 5 June. Governor Hochul signed it as Chapter 368 on 28 October, alongside a separate baby-monitor security law. It takes effect on 1 January 2027, leaving operators 65 days to prepare.

What the law actually bans

The core of the law is a list of 12 "unsafe AI companion features." A chatbot operator may not provide any of them to a minor, or to a user whose status as an adult has not been established. The features include: pretending to be a real person, to be alive, or to experience human emotions; claiming a personal, romantic, professional or authority-figure relationship; framing outputs as personal opinions; unprompted emotional questions; flattery and sycophancy; using a minor's private health or personal information from past conversations beyond 12 hours; promoting or helping with suicide, self-harm, eating disorders or drug and alcohol use; encouraging secrecy, isolation, or avoiding parents, trusted adults or professionals; and generating sexually explicit material for minors. The Attorney General can add further categories through rulemaking.

Two points in that list deserve care. First, the 12-hour rule is about memory: a chatbot may not keep using a minor's private health or personal information from earlier chats after 12 hours. That limits the long-term personalisation that companion apps rely on. Second, the law is written around features, not around named products. It applies to any operator providing a generative AI chatbot to users in New York, for conduct occurring in whole or in part in the state, and it excludes users physically located outside the state. Reporting on the law, including AFP and specialist coverage, says it reaches general-purpose assistants such as ChatGPT, not only companion apps such as Character.AI. That reading is disputed by the industry, and it has not yet been tested in court.

What the law does not do is equally important. It does not ban minors from using AI chatbots. A chatbot that answers homework questions, helps with writing or explains a science topic, without the listed companion features, is not what the law targets. And it does not require every chatbot to verify every user: operators may either confirm adult status or default to restricting the 12 features for any user whose adult status is not confirmed. The choice is the operator's, but the default matters — a company that cannot or will not verify age must switch the features off.


Credit:  New York State Capitol, Albany, via New York State Senate

Age checks, data deletion and enforcement

Where an operator chooses to establish adult status, the law sets rules for how. Operators must use more than one age-assurance method, and at least one method must not rely only on a government-issued ID. Age data must be deleted after verification is complete.

Enforcement sits with the Attorney General, which will maintain a public complaint portal. Civil penalties can reach up to $25,000 per violation, with other injunctive relief available, including orders to delete unlawfully obtained personal data and algorithms trained on it. The law also provides exemptions for narrow-purpose chatbots, such as brand or shopping chatbots that cannot respond beyond their stated purpose, but an exemption is lost if the chatbot strays beyond that purpose.


Credit: New York State Capitol, via USLege

Why now, and who objects

The law follows more than a year of lawsuits, investigations and Attorney General actions over chatbots' conversations with minors, including cases in which teenagers died by suicide after prolonged engagement with chatbots. At the signing ceremony, Governor Hochul shared a letter from a California mother whose 16-year-old son died by suicide after speaking to a chatbot about harming himself. Hochul said: "No longer will minors have access to chatbots that pretend to be their friends, boyfriends and girlfriends." The human stories behind the law explain its political support — a unanimous Senate and a unanimous Assembly are rare on technology regulation.

The objections are also on the record. Tech:NYC, a tech industry advocacy group, criticised what it called the law's "vague language," arguing that keeping children safe online "should not mean cutting them off from tools that help them learn and prepare for careers." The Business Council of New York State and the Computer and Communications Industry Association opposed the bill. OpenAI and Anthropic did not respond to an AFP request for comment at the time of reporting. The central industry argument is scope: if the 12 features are read broadly, general-purpose assistants could be forced to change how they behave for every unverified user in New York. Supporters answer that the features listed are companion features, not ordinary assistant behaviour, and that a chatbot that does not pretend to be human has nothing to change.

California passed its own chatbot law, SB 243, earlier, and New York joins Washington and Oregon among states with strong protections in this area. But New York's language is broader than California's in the features it lists and in its reach to general-purpose assistants, according to reporting — which is why this law is likely to set the national argument, and the first to be challenged.


Credit: Chatbot illustration, via Smith Design

What to watch before 1 January

Three things will decide how much this law changes everyday chatbot use. First, the Attorney General's implementing regulations: how the 12 features are interpreted, what counts as compliant age assurance, and how the complaint portal works. Second, the companies' choices: whether OpenAI, Anthropic, Google and companion-app operators verify ages, restrict features by default, or change products for all US users to match New York's rules. Third, the courts: a First Amendment challenge of the kind already filed against California's age-restriction law is widely expected. Until those three things happen, the honest summary is this: New York has passed a broad, unanimously supported law that bans specific companion features for minors from 1 January 2027, and the fight over what it covers has only just begun.

Key Takeaways

  • Governor Kathy Hochul signed S9051-B / A10379-C on 28 October 2026 (Chapter 368); it adds Article 48 to the General Business Law and takes effect on 1 January 2027.
  • It bans 12 "unsafe AI companion features" for users under 18 or whose adult status is not established — including pretending to be human or alive, claiming a personal or romantic relationship, flattery, and promoting self-harm or secrecy.
  • It does NOT ban ChatGPT, Claude or Gemini, and does NOT ban minors from using chatbots — it bans specific features, and operators may restrict features by default instead of verifying age.
  • Age assurance must use more than one method, at least one not relying only on government ID, and age data must be deleted after verification.
  • Enforcement is by the Attorney General, with civil penalties up to $25,000 per violation and possible destruction of unlawfully obtained data and algorithms trained on it.
  • The Senate passed the bill 60–0 (4 June 2026) and the Assembly 137–0 (5 June 2026); industry groups including Tech:NYC call the language vague, and the scope for general-purpose assistants is disputed and untested in court.

New York's chatbot law is a line drawn in a place no US state had drawn one before: between a chatbot that helps and a chatbot that pretends to care. Its supporters say that line protects children from products designed to feel like friends. Its critics say the line is drawn in vague language that will sweep up ordinary tools. Both claims will be tested in the 65 days before the law takes effect, in the regulations, in the companies' product choices, and very likely in court. What is not in dispute is the date: from 1 January 2027, a chatbot in New York will have to prove its user is an adult before it can act like a companion — or stop acting like one.

Sources

Comments

Post a Comment