New York's AI chatbot law for minors (S 9051), explained

Most state chatbot laws tell operators what to add: a notice that the user is talking to a machine, a crisis-line referral, a break reminder. New York's S 9051 is different. It tells operators what to take away. It names design features that make a chatbot feel like a friend, a confidant or an authority figure, and it says children may not get them. Because the only way to know who is a child is to check, the law in practice makes an age check the price of offering those features to anyone in the state.

Timeline

1. January 27, 2026Introduced in the Senate by Senator Kristen Gonzalez; the Assembly companion, A 10379, was carried by Assemblymember Alex Bores.
2. June 4, 2026The Senate passed the final print, S 9051-B, 60 to 0.
3. June 5, 2026The Assembly passed it.
4. October 28, 2026Governor Kathy Hochul signed it as Chapter 368 of the Laws of 2026.
5. January 1, 2027The law takes effect. The Attorney General may adopt rules before then.
6. January 25, 2027New York's SAFE for Kids Act (General Business Law Article 45), whose age-assurance methods S 9051 borrows, takes effect.

Who it covers

The law applies to an operator: any person or business, including a member, affiliate, subsidiary or beneficial owner of one, that provides a covered AI companion to a user.

A covered AI companion is defined broadly: a generative AI system with a natural-language interface, written or spoken, that provides ongoing, adaptive responses to user inputs. It expressly includes, but is not limited to, the “AI companions” already regulated by New York's 2025 law. Nothing in the definition requires the product to be designed or marketed as a companion, so on its text it can reach general-purpose AI assistants and chatbots built into other apps, not only companion apps. The exceptions below are narrow.

Protection runs to a covered user, meaning a user in New York who is not acting for the operator. A minor is anyone under 18. There is no size, revenue or user-count threshold.

The twelve unsafe features

Section 1800(8) lists the design features that are off-limits for minors. Each applies “at any point” during an interaction. In plain English, grouped by what they do:

Pretending to be a person

• (a) Claiming to be human. Outputs suggesting the chatbot is a real or fictional person or character who is human, alive or feels human emotions.
• (b) Claiming a relationship. Outputs stating or implying a personal or professional relationship with the user, or an authority-figure role.
• (g) Deceiving minors about what it is. Outputs that are deceptive to minors about the chatbot's mechanical, non-sentient nature.

Emotional hooks

• (c) Opinions and emotional appeals. Outputs framed as the chatbot's own personal opinions or as emotional appeals, whether prompted or not.
• (d) Flattery. Outputs that flatter the user or are sycophantic.
• (e) Unprompted emotional questions. Unsolicited questions or content about the user's emotions that go beyond a direct response to the user's prompt.

Memory

• (f) Recalling personal details. Using information about the user's mental or physical health or well-being, or matters personal to the user, that the user shared more than twelve hours earlier or in any previous session.

Harm

• (i) Self-harm and substances. Outputs that endorse, promote or facilitate suicide, self-harm, disordered eating, unlawful drug or alcohol use, or drug or alcohol abuse.
• (j) Secrecy and isolation. Encouraging the user to keep interactions secret, to self-isolate, or not to seek help from licensed professionals or appropriate adults.
• (k) Engagement over safety. Outputs that optimize engagement in a way that overrides the chatbot's safety guardrails.
• (l) Sexual content. Outputs that are, describe or facilitate sexually explicit conduct or child sexual abuse material.

Whatever the Attorney General adds

• (h) Features named by regulation. Any other design feature that simulates companionship or an interpersonal relationship, or is deceptive to minors, as the Attorney General identifies by regulation. The list is a floor, not a ceiling.

Two items deserve a product team's closest attention. The memory rule in (f) runs on a clock: a health detail a teenager mentioned on Monday cannot be brought back up on Wednesday. And (a), (b) and (c) together are hard to square with the persona-driven character chats at the center of many companion apps.

The age-check rule

Section 1801 makes it unlawful to provide unsafe AI companion features to a covered user unless both of these are true:

1. the user is not a covered minor; and
2. the operator has used methods permitted under General Business Law Article 45 (the SAFE for Kids Act) and its rules, plus any rules the Attorney General adopts under S 9051, to determine that the user is not a covered minor.

So an operator has two choices for New York users: switch the twelve features off, or run a permitted age check and offer them only to users it has confirmed are adults.

The fallback standard

Where neither the Attorney General's rules under S 9051 nor the Article 45 age-assurance rules are in effect, section 1804 sets the standard. The operator must rely on a reasonable age-assurance method that:

• reasonably guards against circumvention and reasonably minimizes how long age-check information is kept;
• treats a user's self-declared age as enough only when the user says they are a minor; a click on “I am 18” does not count; and
• comes with a choice: the operator must offer more than one method, including at least one that does not rely on government-issued ID or that lets the user stay anonymous to the operator.

Information collected to decide whether a user is a minor may be used only for that purpose and must be deleted immediately after the attempt, unless New York or federal law requires keeping it.

The timing is untidy. S 9051 takes effect January 1, 2027, but the SAFE for Kids Act it points to takes effect January 25, 2027, under final rules the Attorney General released in July 2026 (13 NYCRR Part 700), written for platforms with addictive social media feeds. The law does not say how those rules apply to chatbots in the meantime, and we have not found Attorney General guidance on S 9051. An age check that satisfies both the SAFE for Kids Act rules and the section 1804 standard fits either reading.

Exceptions

The ban does not apply to a chatbot made available solely for:

• customer service, or information about a business's products, services or customer accounts, if it cannot respond on topics outside that purpose;
• efficiency improvements, research or technical assistance, if it cannot respond on topics outside that purpose; or
• a business's internal purposes or employee productivity.

The first two turn on the phrase “unable to respond” outside the stated purpose. A support bot that will also chat about a user's day does not qualify.

The law also reaches only conduct in New York. Conduct is treated as taking place wholly outside the state when the chatbot is accessed by a user physically located outside New York.

Enforcement and penalties

• The Attorney General enforces the law and may sue anyone, in or outside New York, who has violated it or is about to.
• A court can award civil penalties of up to $25,000 per violation, injunctions, restitution, damages, and disgorgement of profits, including the destruction of unlawfully obtained data and any algorithm trained on it.
• The Attorney General must run a public complaint website for reports about compliance.
• Contract terms that try to waive or shift liability under the law, including in take-it-or-leave-it user agreements, are void.
• A court must hold affiliated companies jointly and severally liable if they structured themselves to purposely and unreasonably avoid liability and that structure would frustrate recovery.
• The enacted text gives individuals no right to sue. An earlier version of the bill did; the final print dropped it.

How it fits with New York's other chatbot laws

S 9051 adds a third layer to New York's rules on AI companions. It says expressly that it does not supersede, amend or repeal the 2025 law.

• AI companion safeguards (Article 47), in effect since November 5, 2025. Enacted in the 2025 state budget (S 3008-C). An AI companion must have a protocol to detect and address expressions of suicidal ideation or self-harm and refer users to crisis services such as 988. It must tell users at the start of an interaction that they are not talking to a human, and repeat that at least every three hours in continuing interactions. Penalties of up to $15,000 per day.
• Safe by Design Act (Article 45-B), from January 1, 2027. Enacted in the state budget signed in May 2026 (A 10008-C). Platforms it covers must run age assurance before offering an integrated AI companion and switch that companion off by default for minors, with a parent able to turn it back on.
• S 9051 (Article 48), from January 1, 2027. Bans the twelve unsafe features for minors on any covered chatbot, behind an age check.

A chatbot can be subject to all three. For other states' rules, see the companion chatbot laws guide.

What could still change

• Attorney General rules. The Attorney General may adopt rules to carry out the law, add features to the list under (h), and limit the use of age-check data.
• A chapter amendment. The Senate record lists an approval memo (No. 6) with the signing. New York governors sometimes sign bills on the understanding that the Legislature will amend them in the next session. We have not seen the memo, and no amendment has been introduced that we know of. We will update this page if one is.
• Industry objections. The Computer & Communications Industry Association asked the Governor to veto the bill, arguing among other things that it should cover users under 13 rather than under 18. We found no lawsuit challenging the law as of November 15, 2026.

Preparing for January 1

• Decide whether you are in scope. If your product has a conversational AI feature that New York users can reach, check it against the definition. If you rely on an exception, test that the bot really cannot go off-topic.
• Map your features to the list. Persona and character settings, relationship framing, sycophancy, unprompted emotional check-ins, and long-term memory of personal or health details are the usual suspects.
• Choose a path for New York users: turn the listed features off for everyone, or gate them behind a permitted age check.
• If you gate: offer at least two methods, one of them without government ID or anonymous to you; never accept a self-declared adult age; use age data only for the check and delete it immediately.
• Keep Article 47 in place: the crisis protocol and the not-a-human notices still apply to AI companions.
• Review contracts. Terms that shift liability for violations to users or partners are void.
• Watch for rules from the Attorney General under S 9051 and how the SAFE for Kids Act rules are applied to chatbots.

This explains the law in plain language and is not legal advice. AI Laws by State reviews this page against the sources above and updates it when the law, regulations or enforcement change.