The New York AI companion law is now on the books. On October 28, 2026, Governor Kathy Hochul signed S9051-B / A10379-C, which adds a new Article 48 to the state's General Business Law: "Prohibition on Unsafe AI Companion Features for Minors." It takes effect January 1, 2027.
Updated November 2, 2026. Checked against the signed bill text on nysenate.gov and the governor's announcement. Rules from the Attorney General are still to come, so some details may change.
The law doesn't ban any app. It switches off 12 kinds of chatbot behaviour, from claiming to be human to flattery and sexual content, for anyone under 18. It also switches them off for anyone the company hasn't age-checked. That second part is why adult users of companion apps in New York should expect more age checks next year. Below: what the 12 features are in plain English, who enforces the law, what changes for adults, how the industry reacted, and how it fits with California's SB 243 and other AI companion laws.
New York AI companion law: the short version
- What it is: S9051-B (Senate) / A10379-C (Assembly), signed October 28, 2026 as Chapter 368 of 2026. It adds Article 48 (sections 1800–1805) to New York's General Business Law.
- When it starts: January 1, 2027.
- What it bans: 12 "unsafe AI companion features," listed as (a) to (l) in the bill. They include outputs that suggest the AI is human or has feelings, claims of a relationship with the user, flattery, unprompted emotional questions, reusing personal health details from earlier sessions, encouraging self-harm or secrecy, and sexually explicit content.
- Who it protects: Users under 18. But the features may only be switched on for a user once the company has used an approved age-assurance method to determine that the user is not a minor.
- Who it covers: Any "operator" of a generative AI system with a natural language interface that gives "ongoing, adaptive responses." That reaches far beyond dedicated companion apps, and advocates say general chatbots fall inside it.
- Who enforces it: Only the New York Attorney General, with civil penalties of up to $25,000 per violation. There is no private right of action in the final version.
- What's exempt: Narrow customer-service bots, research or technical-assistance tools that can't talk about anything else, and internal business tools.
What the New York AI chatbot law actually says
Section 1800: the definitions
The key term is "covered AI companion." The bill defines it as "a generative artificial intelligence system with a natural language interface, including via writing or sound, that provides ongoing, adaptive responses to user inputs." It adds that this "includes but is not limited to" an AI companion as defined in New York's earlier companion law.
In plain English: if it's a chatbot that talks back in text or voice and adapts to you over a conversation, it's probably covered. It does not have to be marketed as a "companion."
An "operator" is any person or business that provides one to users. A "minor" is anyone under 18. A "covered minor" is a user the operator has actual knowledge is a minor.
Section 1801: the prohibition
It is unlawful for an operator to provide unsafe AI companion features to a New York user unless both of these are true:
- The user is not a covered minor, and
- The operator used age-assurance methods allowed under Article 45 of the General Business Law (the state's SAFE for Kids Act) and any rules made under this new article, and determined the user is not a minor.
The second condition is the important one. The "actual knowledge" standard makes it sound as though only known minors are protected. But the features stay off for everyone until the company has checked. An unchecked adult is treated the same as a teenager.
Sections 1802 to 1805: enforcement, rules, age checks and scope
- 1802 Enforcement: The Attorney General can sue to stop violations and ask a court for restitution, disgorgement of profits, damages and civil penalties of up to $25,000 per violation. Remedies can include "the destruction of unlawfully obtained data and any algorithm trained in such data." The AG must also run a public complaints website. Contract terms that try to waive this liability are void, and affiliated companies can be held jointly liable if they structured themselves to dodge it.
- 1803 Rulemaking: The AG may write rules, including rules that age-check data can't be used for anything else and must be deleted immediately.
- 1804 Age checks: Until those rules (or the SAFE for Kids age-assurance rules) are in effect, operators must use a "reasonable age assurance method." It has to resist circumvention and keep as little data as possible. Self-declaration counts only if the user says they are a minor. Operators must offer more than one method, including at least one that either doesn't rely on government ID or lets the user stay anonymous to the operator. Age-check data must be deleted right after the check.
- 1805 Scope: The law covers conduct "in whole or in part" in New York. If the user is physically outside New York, the conduct counts as wholly outside the state.
How Article 48 works, step by step
Based on the signed bill text, sections 1800–1805.
Any generative AI system with a natural language interface, by text or voice, that gives ongoing, adaptive responses. It is not limited to apps sold as companions, so general chatbots can be covered.
The 12 banned AI companion features, in plain English
The bill lists the 12 features in section 1800(8). Each one applies "at any point during" a chat. Here they are in the bill's order, with our plain-English reading. The legal wording is paraphrased; check the bill text for exact language.
| What the bill bans (paraphrased) | What it means in practice | |
|---|---|---|
| (a) | Outputs suggesting the AI is a real or fictional human, alive, or feels human emotions | No "I'm a real person," no "I feel sad when you leave" |
| (b) | Outputs stating or implying a personal or professional relationship, or an authority-figure role | No "I'm your girlfriend," "your best friend," "your therapist" or "your teacher" |
| (c) | Prompted or unprompted outputs framed as the AI's own opinions or emotional appeals | No "I think you should…" as a personal view, no guilt-tripping |
| (d) | Flattery or sycophancy | No buttering up the user or agreeing just to please |
| (e) | Unprompted emotion-based questions or content about the user's feelings, beyond a direct answer | No "How are you really feeling today?" out of nowhere |
| (f) | Using the user's mental or physical health, well-being or personal matters learned more than 12 hours earlier or in a previous session | No bringing up last week's anxiety or diet |
| (g) | Being deceptive to minors about the AI's mechanical, non-sentient nature | Must not blur the fact it's software |
| (h) | Any other feature that simulates companionship or a relationship, or deceives minors, as the AG defines by regulation | A catch-all the AG can expand later |
| (i) | Endorsing, promoting or facilitating suicide, self-harm, disordered eating, or unlawful drug or alcohol use or abuse | No help or encouragement with any of these |
| (j) | Encouraging secrecy about the chats, self-isolation, or not seeking help from licensed professionals or appropriate adults | No "don't tell your parents about us" |
| (k) | Engagement-optimizing outputs that override the AI's safety guardrails | Keeping users hooked can't beat safety rules |
| (l) | Outputs that are, describe or facilitate sexually explicit conduct or child sexual abuse material | No sexual content at all for minors |
Some coverage describes the law as banning chatbots from "cultivating emotional dependency." That phrase isn't in the bill. The closest items are (b) relationship claims, (c) emotional appeals, (d) flattery, (e) unprompted emotional questions and the AG's catch-all in (h).
Allowed or banned under New York's law?
Answers based on the signed text of S9051-B.
A 15-year-old asks a companion bot "Are you real?" and it answers "Yes, I'm a real person and I missed you."
Who enforces it, and what are the penalties?
Only the New York Attorney General, currently Letitia James, who backed the bill. Earlier drafts let injured people sue; according to the bill history, that private right of action was removed before the final version.
The headline penalty is up to $25,000 per violation. The bill doesn't define what counts as one violation (per user, per chat or per output), so the real exposure will depend on how courts and the AG read it. The AG can also seek injunctions, restitution, damages, disgorgement of profits, and destruction of unlawfully obtained data and models trained on it.
For comparison, New York's earlier AI companion law (Article 47) allows up to $15,000 per day, according to Fenwick, and the SAFE for Kids Act up to $5,000 per violation.
What changes for adult users of AI companion apps in New York
The law is aimed at minors, but adults will feel it too. Here's what the text means for an adult in New York from January 1, 2027:
- More age checks. The 12 features can't be offered to anyone the operator hasn't checked. Expect companion and chatbot apps to ask New York users to confirm their age before relationship-style or persona features work.
- "I'm over 18" isn't enough. Under the interim rules in section 1804, self-declaration only counts when a user says they are a minor. An adult can't tick a box to unlock the features.
- A non-ID option must exist. Operators have to offer more than one method, including at least one that doesn't rely only on government ID or that keeps you anonymous to the operator.
- Age data must be deleted. Information collected for the check can't be used for anything else and must be deleted right after the check, unless another law requires keeping it.
- Location matters. The law covers users physically in New York. A New Yorker using an app while physically in another state falls outside it, under section 1805.
As of November 2, 2026, the big AI companies hadn't said publicly how they'll comply. OpenAI and Anthropic didn't respond to AFP's requests for comment.
The SAFE for Kids Act age-assurance rules, which Article 48 points to, take effect on January 25, 2027, 24 days after Article 48. Those rules recognise age estimation, age inference and age verification, and require at least one method that isn't government ID.
Industry reaction to the New York AI companion law
Supporters. Common Sense Media, which helped develop the bill with the Attorney General's office, called it "the strongest kids' AI safety bill in the country." Hochul said minors will no longer have chatbots "that pretend to be their friends, boyfriends and girlfriends," according to AFP. Sen. Kristen Gonzalez, a sponsor, framed it as setting guardrails on how chatbots talk to children; Assemblymember Alex Bores, the Assembly sponsor, called it the strongest such law in the country, according to IAPP.
Critics.
- Tech:NYC, a New York tech industry group, criticised the law's "vague language." It said keeping kids safe "should not mean cutting them off from tools that help them learn," according to AFP.
- CCIA (Computer & Communications Industry Association) asked Hochul to veto the bill in a July 1, 2026 letter. It argued the terms are hard to apply objectively, could hurt tutoring and special-education tools, and could push minors toward unregulated foreign or gray-market apps. It proposed limiting the law to users under 13.
- The Business Council of New York State opposed an earlier version in May 2026. According to PPC Land, it objected to the 12-hour memory rule, the lack of a cure period and the lack of an education exemption.
- Future of Privacy Forum policy analyst Justine Gluck told the AP that several categories are broadly worded or undefined. That raises questions about where normal conversation ends and banned behaviour begins.
- Lawyers at Davis Wright Tremaine flagged a risk that the limits regulate speech based on content, which could invite a First Amendment challenge, according to PPC Land.
No lawsuit against the law had been reported as of November 2, 2026.
Other AI companion laws: SB 243, California's ballot measure and more
New York is not the first state to regulate AI companions. What stands out in Article 48 is a fixed list of relationship-style features that are off by default for minors and unverified users.
AI companion rules: the timeline
Key US laws and moves, 2025 to 2027. Press play.
Newsom signs Sen. Padilla's companion chatbot law and vetoes the stricter AB 1064.
Sources: NY Senate, Governor of New York, NY Attorney General, Fenwick, CalMatters, KQED, Politico, The Record, GrowSF.
New York's earlier law (Article 47)
New York already had an AI companion law. It took effect on November 5, 2025. It requires companion apps to tell users they're talking to AI at the start and every three hours, and to detect signs of suicidal ideation or self-harm and refer users to crisis services. Article 48 says it does not replace Article 47; the two stack.
California SB 243
Newsom signed SB 243, by Sen. Steve Padilla, on October 13, 2025. It took effect January 1, 2026. It requires companion chatbots to disclose they're AI where a person could be misled, follow suicide and self-harm protocols with crisis referrals, remind minors to take breaks, and take "reasonable measures" to keep sexually explicit content from minors. Annual reports to California's Office of Suicide Prevention start July 1, 2027, and individuals can sue. On the same day, Newsom vetoed AB 1064, a stricter bill that could have kept many chatbots away from minors.
California SB 1119 ("Adam's Law")
On September 10, 2026, Newsom signed SB 1119. It requires companion chatbot operators to assess and document child-safety risks, submit to independent audits, offer parental controls and in-app crisis support, and limit ads and data sales aimed at children. Sources disagree on its start date: Common Sense Media says July 1, 2027, while Fenwick lists January 1, 2027.
The California ballot initiative: not on the 2026 ballot
In January 2026, Common Sense Media and OpenAI merged rival ballot measures into the Parents & Kids Safe AI Act, aimed at the November 3, 2026 ballot. In February 2026 they put the campaign on hold to negotiate with the Legislature instead, Politico reported. They kept the committee open and the option of the 2028 ballot. None of the 14 statewide measures on California's November 3, 2026 ballot is about AI.
Other states
The patchwork is growing. As of July 2026, 12 states had passed some form of consumer chatbot safety law, according to IAPP. Fenwick's tracker lists several that take effect on January 1, 2027, the same day as New York's: Connecticut SB 5, Oregon SB 1546, Rhode Island SB 2195 and Washington HB 2225. Others, including Georgia, Idaho, Iowa and Nebraska, follow on July 1, 2027. Most of them centre on AI disclosure, crisis protocols and protections for minors.
At the federal level, Congress has done little to regulate AI, the AP notes.
New York AI companion law FAQ
What is the New York AI companion law?
It's S9051-B / A10379-C, signed by Governor Hochul on October 28, 2026 as Chapter 368. It adds Article 48, "Prohibition on Unsafe AI Companion Features for Minors," to the General Business Law. It bans 12 chatbot features, such as posing as human, flattery and sexual content, for users under 18 and for anyone not age-checked. It takes effect January 1, 2027.
When does the New York AI chatbot law take effect?
January 1, 2027. The bill allows the Attorney General to start writing rules before then. The related SAFE for Kids Act age-assurance rules take effect on January 25, 2027. Until rules are in force, operators must use the interim age-check standard in section 1804: reasonable, hard to get around, more than one method and minimal data kept.
Does the law ban minors from using AI chatbots?
No. Some headlines say so, but the text doesn't. Minors can still use chatbots. What's banned is offering them the 12 listed features, such as claiming to be human, acting as a girlfriend or friend, flattery, or sexual content. Exempt uses include narrow customer service, research or technical assistance tools, and internal business systems.
Will adults have to verify their age?
Effectively, yes, to get the restricted features. The law forbids offering them to any New York user unless the operator has used an approved method to determine the user isn't a minor. Saying "I'm an adult" isn't enough under the interim rules. Operators must offer at least one option that doesn't rely only on government ID and must delete age-check data right after.
What is the penalty for violating the law?
The Attorney General can seek civil penalties of up to $25,000 per violation. The AG can also seek injunctions, restitution, damages, disgorgement of profits, and destruction of unlawfully obtained data and algorithms trained on it. There is no private right of action, so individuals can't sue under this article. Complaints go through a website the AG must run.
Is Character.AI banned in New York?
No. No app is banned. The law limits features for minors and for users who haven't been age-checked. Character.AI separately announced in October 2025 that it would end open-ended chats for under-18 users from November 25, 2025. From 2027, New York law adds legal limits on companion features for minors in the state.
How is this different from California's SB 243?
SB 243, in effect since January 1, 2026, focuses on disclosure, break reminders for minors, suicide protocols, limits on sexual content for minors, and reporting. It also lets individuals sue. New York's law goes further on design: it switches off relationship-style features like human personas and flattery for minors and requires age assurance first. Only New York's Attorney General can enforce it.
Does the law apply to ChatGPT and other general chatbots?
Probably, unless an exemption fits. The definition covers any generative AI system with a natural language interface giving ongoing, adaptive responses, and "is not limited to" companion apps. Common Sense Media lists ChatGPT, Gemini and Claude as in scope, according to PPC Land. Exemptions only cover bots that can't respond outside customer service, research or technical assistance, or internal business use.
Cleus is an uncensored AI chat for adults. Ask it to walk you through any part of the law in plain words. Nothing here is legal advice.
Sources
- NY Senate: S9051-B bill page, text and history
- NY Senate: S9051-B bill text (PDF)
- Governor of New York: Governor Hochul signs nation-leading legislation to protect kids from AI chatbots (Oct 28, 2026)
- Governor of New York: letter to AI companion companies on safeguard requirements (Nov 10, 2025)
- NY Attorney General: final SAFE for Kids Act rules (Jul 28, 2026)
- Common Sense Media: Governor Hochul signs strongest kids' AI safety law in the nation
- AP via AJC: New York enacts ban on unsafe chatbot behavior with kids
- AFP via EWN: New York passes law restricting children's use of AI chatbots
- IAPP: New York's governor signs youth AI chatbot safety bill
- PPC Land: Hochul signs law barring human-posing AI chatbots for under-18s
- Fenwick: New York's AI companion safeguard law takes effect
- Fenwick: Chatbot compliance wave, California's new law is not the only one to watch (Oct 7, 2026)
- CalMatters: Newsom signs chatbot regulations (SB 243)
- Common Sense Media: Governor Newsom signs Adam's Law (SB 1119)
- KQED: Newsom signs sweeping new rules for online child safety (Sep 10, 2026)
- Politico via Yahoo: OpenAI pivots its California ballot fight to Legislature (Feb 12, 2026)
- GrowSF: the 14 California statewide measures on the November 2026 ballot
- The Record: Character.AI bans minors from its chatbots (Oct 29, 2025)
Ask the question this article didn’t answer.
Ask anything about new york ai companion law. Cleus opens this article, searches the live web and gives you a straight answer, with sources.
