Starting Jan. 1, a chatbot talking with a New York teenager is supposed to forget what she told it. If a 15-year-old in Syracuse mentions on a Sunday night that she is dreading a meeting with her guidance counselor, state law will bar the bot from raising it again on Monday afternoon. Under the new rule, a chatbot may not use anything personal, or anything about her mental or physical health, that it learned more than 12 hours earlier or in a previous session.

That rule is one of 12 written into Chapter 368, the bill Gov. Kathy Hochul signed on Oct. 28 after it cleared the Senate 60 to 0 on June 4 and the Assembly a day later. Carried by Sen. Kristen Gonzalez as S9051B and by Assemblymember Alex Bores as A10379C, it adds a new Article 48 to the General Business Law and takes effect on Jan. 1, 2027. From that day, companies may not offer any of the 12 listed unsafe AI companion features to a New York user under 18, or to any user they have not first determined is an adult.

Twenty-four days later, on Monday, Jan. 25, the SAFE for Kids Act starts restricting algorithmic feeds and overnight notifications on social media for the same age group. The two laws arrive on the same phones in the same month, and they are built differently enough that a family will want to know which one governs which app before the holiday break ends.

At the signing, Hochul described the first law in terms any parent of a teenager would recognize. "No longer will minors have access to chatbots that pretend to be their friends and boyfriends and girlfriends. No longer will chatbots be allowed to keep sensitive information about a minor's emotional and physical health. No longer will chatbots discourage kids from seeking help from licensed professionals," she said, according to the governor's transcript.

The 12 Features Leaving Teen Chats

The first eight items on the list describe a personality. A covered chatbot may not suggest to a minor that it is a real or fictional character who is human, alive or capable of human emotions. It also may not state or imply that it holds a personal, professional or authority-figure role in the user's life. That second clause reaches past the girlfriend apps to the tutor, coach and counselor personas a teenager might choose for homework help or a hard week.

Opinions and emotional appeals framed as the chatbot's own are barred whether the user asked for them or not. Flattery and sycophancy are out, along with unsolicited questions about the user's feelings that go beyond answering what was asked. The 12-hour memory limit sits in this group, next to a ban on outputs that mislead minors about the mechanical, non-sentient nature of the product. A catch-all lets the attorney general add other companionship features by regulation.

The last four items target outcomes rather than personality. A chatbot may not endorse, promote or facilitate suicide, self-harm, disordered eating, or unlawful drug or alcohol use. Encouraging a minor to keep the conversation secret, to pull away from people, or to avoid licensed professionals and appropriate adults is prohibited, as is chasing engagement in ways that override the product's own safety guardrails. Sexually explicit material and child sexual abuse material close the list. In the signing announcement, Hochul put the purpose in one line: "No company should be peddling chatbots to kids encouraging self-harm or pretending to be their friend or more."

The Definition Reaches Past Companion Apps

The statute's definition is wider than its nickname. A covered AI companion is any generative AI system with a natural-language interface, written or spoken, that gives ongoing, adaptive responses to what a user types or says. The text specifies that the term includes but is not limited to the products New York law already calls AI companions.

Only three kinds of chatbots are exempt. Customer service bots qualify if they cannot respond on topics outside that purpose, and so do tools built for efficiency, research or technical assistance under the same restriction. Systems a business uses internally or for employee productivity are also outside the law. A general-purpose assistant that will walk a 16-year-old through a chemistry lab at 4 p.m. and a breakup at midnight fits none of those categories. Common Sense Media made the bill its top New York priority this year. In its statement on the signing, the group named ChatGPT, Gemini, Meta AI, Snap's My AI, Character.AI and Replika among the chatbots the law is meant to cover.

New York already has one AI companion law on the books. Since Nov. 5, 2025, companion products have had to detect expressions of suicidal thinking and direct those users to crisis services. They must also tell users at the start of a session, and every three hours after that, that they are not talking to a person. Family Symposium walked through those duties in its guide for parents of teens using AI companions. Article 48 states that it does not supersede, amend or repeal that law, so the crisis referral duty continues for users of every age while the new feature ban sets a separate floor for minors.

How a Chatbot Is Supposed to Know Who Is 17

The law puts the burden on the company rather than the family. Under Section 1801, offering any of the 12 features to a New York user is unlawful unless the user is not a known minor and the operator has used approved methods to establish that. A company that never checks has to treat every user as if the restrictions apply.

Until the attorney general writes formal rules, Section 1804 sets interim standards for age assurance. Any method must reasonably guard against circumvention and keep as little of the collected information as possible. Self-declaration works in only one direction, since an operator may rely on it when a user says she is a minor, while a 14-year-old who types in an adult birth year has proved nothing. Companies must offer more than one method, including at least one that either does not rely on a government ID or lets the user stay anonymous to the company. Whatever is gathered for the check must be deleted immediately afterward unless another state or federal law requires keeping it.

Geography sets the outer edge. The article applies to conduct that happens wholly or partly in New York and treats a session as wholly outside the state when the user is physically somewhere else. A 17-year-old texting a chatbot from her grandmother's house in Florida over the holidays falls outside it, and the same teenager at home in Buffalo is covered wherever the company's servers sit.

What Changed Between January and October

The Senate first printed the bill on Jan. 27 with a broader enforcement section. That version let any injured person sue the company and the individuals controlling its compliance. It also created a rebuttable presumption that a chatbot caused or contributed to the injury whenever a minor harmed themself after the bot encouraged it. The private right of action was gone by the May amendment, which left enforcement to the attorney general alone.

Her office can still seek injunctions, restitution, disgorgement of profits, damages, the destruction of unlawfully obtained data and of any algorithm trained on it, and civil penalties of up to $25,000 per violation. Contract terms that try to waive or shift that liability are void, and courts must hold affiliated companies jointly liable if they structured themselves to dodge it.

The earliest drafts in both chambers would also have barred a chatbot from using "I," "my" and "me" to describe itself. The final text dropped the pronoun ban, kept the prohibition on claiming to be human or alive, and replaced a 180-day clock with a fixed effective date of Jan. 1. "For too long, AI companies have profited from rolling out chatbots to underage consumers," Gonzalez said in the governor's announcement. "We passed S9051B/A10379C alongside Assemblymember Bores to put real guardrails on how chatbots talk to children." Bores pitched the result to a national audience: "For everyone wondering whether the public can have any say in determining how AI impacts us, New York again has an answer: Absolutely."

Jan. 1 and Jan. 25 Run on Different Rules

Under the SAFE for Kids Act, covered social media platforms must show users under 18 a feed of accounts they follow instead of an algorithmic one. The Standard laid out the details in its guide to what changes on teen phones Jan. 25. Feed-related notifications must also stop between midnight and 6 a.m. Eastern. A parent can switch those features back on only after the teenager asks and agrees to bring the parent in, and the attorney general can seek up to $5,000 per violation.

The chatbot law has no consent path at all. A parent who wants a 16-year-old to keep a study bot that remembers last week's lessons cannot sign off on it. The only route to the 12 features runs through proof that the user is an adult. The ceiling on penalties is five times higher. Both laws point to Article 45 of the General Business Law, the SAFE for Kids article, for acceptable age-checking methods, so a teenager may meet similar verification screens twice in the same month.

The two dates also close a gap. SAFE for Kids is written around feeds of posts shared by other users, which left companion chatbots outside its reach. Schools set their own rules. In New York City, companion chatbots are barred in every grade under the moratorium announced in September, as The Standard reported in its coverage of the city's high school AI pilot. Most districts elsewhere in the state have no AI policy, which makes the Jan. 1 law the first statewide chatbot restriction written specifically for users under 18.

What Parents and Counselors Can Report

The law directs the attorney general to maintain a complaint website that accepts information, tips and referrals from the public about whether a company is complying. A complaint carries more weight with a screenshot that shows the app's name, the date and time, and the exchange itself. It also helps to note whether the app ever asked the teenager to verify an age, and which method it offered.

School counselors are likely to hear about these exchanges before parents do. A student who says a chatbot told her not to mention their conversations to anyone, or who describes a bot bringing up something she confided days earlier, is describing conduct the statute names. Melinda Person, president of New York State United Teachers, framed the stakes for educators in the governor's release: "Our kids deserve the chance to learn, grow and build real human connections without technology getting in the way."

Attorney General Letitia James, whose office will field those complaints, spoke in the release about what has already happened. "We have seen too many horror stories of chatbots hurting the health of our children, and this new law sets clear legal parameters to prevent that harm and keep kids safe," she said.

Before Jan. 1, the attorney general may issue rules on age methods and add features to the banned list, and the law authorizes that work to be finished ahead of the effective date. Common Sense Media said lawmakers in Michigan and Florida are expected to consider similar bills in 2027. In New York, the first date falls on a Friday, and the second arrives on a Monday, 24 days later.