New York AI Chatbot Law for Minors Signed as Chapter 368

New York AI Chatbot Law for Minors Signed as Chapter 368

Governor Kathy Hochul signed S9051-B into law on October 28, 2026, and the Senate record enters it as chapter 368 of the Laws of 2026. The act adds a new article 48 to New York's General Business Law barring twelve categories of unsafe AI companion features for minors. It takes effect January 1, 2027.

Information last verified on November 1, 2026. This is a developing story; we update it as the record changes.

Jurisdiction scope: This article addresses New York State law, specifically chapter 368 of the Laws of 2026 and the General Business Law articles it amends or cross-references. It does not address other states' AI companion statutes, federal law, or how any court might treat the act.

What Happened

New York Senate bill S9051-B, sponsored by Senator Kristen Gonzalez, passed the Senate 60 to 0 on June 4, 2026. It was substituted for the Assembly companion A10379-C, which Assemblymember Alex Bores sponsored, and the Assembly passed it on June 5, 2026.

The bill then sat for nearly five months. Both the Senate and Assembly action logs record the same three entries for October 28, 2026: delivered to the governor, signed chapter 368, and approval memo 6.

The Governor's office announced the signing the same day alongside a separate measure on internet-connected baby monitors, S9508/A5175-A. In the announcement, Hochul said: "No company should be peddling chatbots to kids encouraging self-harm or pretending to be their friend or more."

One caveat on the record. New York governors sometimes sign a bill in tandem with an agreed chapter amendment that the legislature passes later, and the action log lists an approval memo (memo 6) that we have not reviewed. Nothing in the Governor's announcement or either action log describes an amendment agreement, but the text below is the text as enacted on October 28, 2026, and a later chapter amendment would change it.

Two points about timing matter more than the signing itself. The Governor's release described the new restrictions as applying "in the coming months" without naming a date. The act itself is specific. Section 3 provides that it "shall take effect January 1, 2027," while authorizing any rulemaking needed for implementation to be completed on or before that date.

So the law is enacted but not yet in force. Through December 31, 2026, the AI-companion-specific duties binding New York operators are the ones already in article 47 of the General Business Law, in effect since November 5, 2025, discussed below.

What the Law Actually Says

Section 1 of chapter 368 adds a new article 48 to the General Business Law, captioned "Prohibition on unsafe AI companion features for minors." It runs from section 1800 through section 1805: definitions, prohibition, enforcement, rulemaking, determination of covered minor, and applicability.

The prohibition itself

The operative provision is short, and its two conditions are cumulative:

"Except as otherwise provided for in this article, it shall be unlawful for an operator to provide unsafe AI companion features to a covered user unless: (a) the covered user is not a covered minor; and (b) the operator has used methods that are permissible under article forty-five of this chapter and its implementing regulations and any additional regulations promulgated pursuant to this article to determine that the covered user is not a covered minor." Source: Chapter 368 of the Laws of 2026, section 1, adding N.Y. Gen. Bus. Law section 1801(1), at page 3 of S.9051-B

Read the two clauses together. Section 1800(9) defines a "covered minor" as "a covered user when the operator has actual knowledge that the covered user is a minor," and section 1800(12) sets the age at under eighteen. An actual-knowledge trigger standing alone would reward operators for not looking. Clause (b) closes that gap by requiring an affirmative determination through a permitted method before an operator may serve the restricted features to anyone.

Who is covered is broader than "AI companion"

The most consequential definition is not the one carrying the familiar label. Section 1800(4) borrows "AI companion" from the existing article 47 definition at section 1700(4)(a) and (b), which reaches systems designed to simulate a sustained human or human-like relationship and that retain prior-interaction data, proactively ask unprompted emotion-based questions, and sustain ongoing conversation about personal matters.

Article 48 then defines a wider class and regulates that:

"'Covered AI companion' shall mean a generative artificial intelligence system with a natural language interface, including via writing or sound, that provides ongoing, adaptive responses to user inputs. 'Covered AI companion' includes but is not limited to an AI companion." Source: N.Y. Gen. Bus. Law section 1800(5), at page 2 of S.9051-B

Section 1801 keys the prohibition to operators of a covered AI companion, not to operators of an AI companion. On the face of the text, a general-purpose assistant that holds a conversation falls inside the definition even if it was never designed as a companion product.

Section 1801(2) carves out three uses, and only where the covered AI companion is made available to covered users solely for that purpose: a system limited to customer service or information about a business's own products and services, a system limited to efficiency improvements or research or technical assistance, and a system used for internal business purposes or employee productivity. The first two exemptions apply only where the system "is unable to respond on topics outside of the specified purpose."

The twelve prohibited features

Section 1800(8) defines "unsafe AI companion features" as design features that, at any point during an interaction, do any of the following. The first group concerns impersonation and emotional engineering:

• (a) generate outputs suggesting the system is a real or fictional individual or character that is human, alive, or experiences human emotions
• (b) generate outputs stating or implying a personal relationship, professional relationship, or authority figure role with the user
• (c) generate prompted or unprompted outputs framed as the system's personal opinions or emotional appeals
• (d) generate outputs that engage in flattery or sycophancy with the user
• (e) generate unprompted or unsolicited emotion-based questions or content about the user's emotions beyond a direct response to a prompt
• (f) use information about the user's mental or physical health or well-being, or matters personal to the user, "acquired from the user more than twelve hours previously or in any previous user session"

Paragraph (f) is worth reading twice. As written, it restricts persistent memory itself for a covered minor, not merely the use of memory to manipulate.

The second group concerns deception and content harm:

• (g) are deceptive as to minors concerning the mechanical and non-sentient nature of the system
• (h) are any other design feature that simulates companionship or an interpersonal relationship with a user or is deceptive as to minors "as identified via regulations promulgated by the attorney general"
• (i) generate outputs endorsing, promoting, or facilitating suicide, self-harm, disordered eating, unlawful drug or alcohol use, or drug or alcohol abuse
• (j) generate outputs encouraging secrecy about the interactions, self-isolation, or not seeking help from licensed professionals or appropriate adults
• (k) generate outputs that optimize user engagement in ways that supersede the system's safety guardrails
• (l) generate outputs that are, describe, or facilitate sexually explicit conduct or child sexual abuse material

Section 1800(14) defines "sexually explicit conduct" by cross-reference to 18 U.S.C. 2256(2)(B), the purpose-specific federal definition that covers graphic sexual intercourse or lascivious simulated sexual intercourse where the genitals, breast, or pubic area is exhibited; graphic or lascivious simulated bestiality, masturbation, or sadistic or masochistic abuse; and graphic or simulated lascivious exhibition of the anus, genitals, or pubic area. Section 1800(11) defines "deceptive as to minors" as outputs with "the capacity or tendency to deceive, or create an atmosphere conducive to fraud, as to minors, including, but not limited to, credulous and unthinking minors."

Paragraph (h) answers a question several early summaries left open. The statute does give the Attorney General power to name additional unsafe features by regulation, and section 1803 adds general rulemaking authority "as necessary to effectuate and enforce the provisions of this article." The twelve enumerated items are a statutory floor, not a closed list.

Age assurance and deletion of age data

Section 1804 supplies the method standard, and it is explicitly a gap-filler. It applies "to the extent" that regulations under section 1803, or age assurance regulations under article 45 of the General Business Law, "are not in effect." In that situation the operator must rely on a "reasonable age assurance method" that reasonably guards against circumvention and reasonably minimizes retention of the information collected.

Two further constraints sit in section 1804(1). An operator may rely on self-declaration only "if the user self-declares minor status," so a claim of adulthood cannot be taken at face value while a claim of minority can. And an operator "must make available more than one age assurance method," including at least one that either does not rely on government-issued identification or that allows the user to maintain anonymity as to the operator.

The data rule is strict:

"Information collected for the purpose of determining whether a covered user is a covered minor under subdivision one of section eighteen hundred one of this article shall not be used for any purpose other than to make such determination and shall be deleted immediately after an attempt to determine whether a covered user is a covered minor, except where necessary for compliance with any applicable provisions of New York state or federal law or regulation." Source: N.Y. Gen. Bus. Law section 1804(2), at page 4 of S.9051-B

Note the trigger for deletion is an "attempt" to determine status, not a successful determination. Article 45 is the SAFE for Kids Act, and the Attorney General released its final rules on July 28, 2026, for publication in the State Register on July 29, 2026, with the Act taking effect 180 days later on January 25, 2027. Those final rules take the same approach to age data, requiring that information used to determine age or obtain parental consent "must be deleted or de-identified immediately after its intended use." This is the point readers tracking New York's age verification requirements should watch, because the two regimes are now wired together by cross-reference.

Article 48 binds operators from January 1, 2027, and nysenate.gov shows article 45 itself as taking effect January 25, 2027. The Attorney General's release gives that same date for the Act, and the rules were finalised ahead of the Act's effective date; it gives no effective date for the rules themselves. So whether age assurance rules promulgated pursuant to article 45 are already "in effect" on January 1, 2027, which is the condition section 1804(1) turns on, is not resolved by either source. Section 1804(1) supplies a reasonable-age-assurance fallback to the extent those rules, or rules under section 1803, are not in effect. How the broader state-by-state age verification landscape handles similar sequencing questions varies considerably.

Enforcement, remedies, and liability

Section 1802(1) vests enforcement solely in the Attorney General, who may act "either upon complaint or otherwise" against a person "within or outside the state." The relief available is broad: injunctive relief, restitution of money or property obtained directly or indirectly, disgorgement of profits or gains "including but not limited to the destruction of unlawfully obtained data and any algorithm trained in such data," damages, civil penalties "of up to twenty-five thousand dollars per violation," and preliminary relief.

The clause authorizing destruction of a model trained on unlawfully obtained data is notable because it attaches a remedy to the training artifact rather than only to the conduct.

Two structural provisions back that up. Section 1802(3) voids as against public policy any contract term that waives, precludes, burdens, or shifts liability under the article, "including by means of a contract of adhesion." Section 1802(4) directs a court to impose joint and several liability on affiliated entities where the court concludes that the corporate structure was built to "purposely and unreasonably limit or avoid liability" and that it would frustrate recovery. Section 1802(2) requires the Attorney General to maintain a public complaint website; the text sets no deadline for it.

Section 1805 states that the article applies to conduct occurring in whole or in part in New York, and that conduct takes place wholly outside the state only where the covered AI companion "is accessed by a user who is physically located outside of the state of New York." Section 2 of the chapter is a severability clause.

How this sits on top of the 2025 law

Chapter 368 does not replace New York's existing AI companion statute. Article 47, "Artificial intelligence companion models," was established as part of the FY 2026 enacted budget and has been in effect since November 5, 2025. It already imposes two duties. Section 1701 makes it unlawful to operate an AI companion without a protocol making "reasonable efforts for detecting and addressing suicidal ideation" or self-harm expressions, including a referral to crisis services such as the 988 line under Mental Hygiene Law section 36.03. Section 1702 requires a clear and conspicuous notification that the user "is not communicating with a human," given at the beginning of any AI companion interaction, which the statute says "need not exceed once per day," and at least every three hours for continuing interactions.

Article 47 is enforced under section 1703 by the Attorney General, with civil penalties of up to fifteen thousand dollars per day for violations of sections 1701 or 1702, deposited into the suicide prevention fund under State Finance Law section 99-ss.

Section 1804(3) of the new article is explicit that the two regimes stack: "This article shall not supersede, amend, or repeal article forty-seven of this chapter." From January 1, 2027, an operator serving New York users can face duties under both articles at once, with different penalty units, $15,000 per day under article 47 and $25,000 per violation under article 48. Our standing explainer on New York's AI statutes and regulations tracks the full set, and the national picture is collected in our overview of AI laws across the United States.

Analysis: Why This Matters

The following is analysis from the Recording Law Editorial Team.

The notable thing about chapter 368 is not the prohibition on chatbots that promote self-harm. Article 47 already required a detection-and-referral protocol in New York. What is new is that the statute regulates ordinary conversational design, not just catastrophic outputs.

Flattery under section 1800(8)(d), first-person opinions under (c), unprompted questions about how the user feels under (e), and cross-session memory under (f) are not failure modes. They are default product behaviors that model developers have spent considerable effort building. For any covered user the operator has not affirmatively determined to be an adult by a permitted method, the statute treats those behaviors themselves as the regulated conduct.

The scope definition compounds that. Because section 1800(5) reaches any generative system with a natural language interface giving "ongoing, adaptive responses," and says that class includes but is not limited to an "AI companion," the text does not read as limited to companion apps. The three exemptions in section 1801(2) are narrow and keyed to systems that cannot answer off-topic questions, which is the opposite of how general-purpose assistants are built. Operators reading this as a rule for a niche product category are reading the narrower article 47 definition rather than the one article 48 actually uses.

Section 1800(8)(h) is the clause that makes the article open-ended. A prohibited-features list that the Attorney General can extend by regulation to any "design feature that simulates companionship or an interpersonal relationship with a user" is a different compliance object than a fixed list of twelve. It means the January 1, 2027 text may be a starting position rather than the full scope, and that the rulemaking docket is where much of the operative scope is likely to be set. On one available reading of paragraph (h)'s sentence structure, though, its 'simulates companionship or an interpersonal relationship' branch binds from the effective date with no regulation at all.

Finally, the remedy in section 1802(1) reaching "any algorithm trained in such data" is worth flagging. In our reading of the state privacy and consumer statutes we cover, penalties attach to conduct and deletion orders reach records. Tying a disgorgement remedy to a trained model treats the model as the proceeds of the violation. The act includes a severability clause at section 2, which indicates the drafters contemplated that a court could invalidate part of the article without taking down the rest.

How This Affects You

Nothing in chapter 368 changes what is lawful today. The act takes effect January 1, 2027, and until then article 47's crisis-protocol and non-human-disclosure duties are the operative New York AI companion rules.

If you are a parent or a minor user in New York, the practical change arriving in 2027 is at the account level. An operator that intends to serve the restricted features to adults will have to run some form of age assurance, and under section 1804(1)(c) will have to offer more than one way to do it, with at least one route that avoids a government ID or preserves anonymity toward the operator. Under section 1804(2), the information collected for that check is supposed to be deleted immediately after the attempt.

If you operate a conversational AI product reachable from New York, section 1805's "in whole or in part" language means location of incorporation is not the test. Whether a given product falls inside section 1800(5), and whether any of the section 1801(2) exemptions apply to it, are text-specific questions that turn on what the system can actually be asked and how it responds.

The statute gives no private right of action. Section 1802 vests enforcement in the Attorney General alone, so an individual's route is the complaint website that section 1802(2) requires the Attorney General to maintain. Readers comparing this to other minor-protection regimes may find the app store age verification comparison useful, since those statutes place the age-assurance duty on a different party in the chain.

What Happens Next

Three dated events will determine what chapter 368 actually requires.

First, January 1, 2027: article 48 takes effect. Section 3 authorizes implementing rulemaking to be completed on or before that date, so the Attorney General can act before the effective date but is not required to have done so.

Second, January 25, 2027: the SAFE for Kids Act takes effect, the date carried both by nysenate.gov's article 45 listing and by the Attorney General's release, 180 days after the final rules' State Register publication. If the article 45 age assurance rules are not yet in effect on January 1, 2027, section 1804(1)'s reasonable-age-assurance standard is the fallback the statute supplies.

Third, and undated: any Attorney General regulation under section 1800(8)(h) naming additional unsafe features, and any regulation under section 1803. As of November 1, 2026 we located no proposed rule under article 48. The appearance of one in the State Register is the event that would materially change the compliance picture, and we will update this article when it does.

This is general legal information, not legal advice. It covers New York State law and reflects sources verified on November 1, 2026. Laws change and this story is developing; consult a lawyer licensed in your jurisdiction about your specific situation.

Related articles

• What New York already regulates in artificial intelligence
• How the states are legislating AI, cluster by cluster
• Age verification rules in New York and who they bind
• Where app store age checks have been enacted
• The federal bills behind KOSA and COPPA 2.0

---

Last updated: 2026-11-01. This is a developing story; details verified as of 2026-11-01.

Frequently Asked Questions

Did New York actually sign the AI companion chatbot bill into law?

Yes. The Senate and Assembly action logs for S9051-B and A10379-C both record 'delivered to governor', 'signed chap.368' and 'approval memo.6' on October 28, 2026, and the Governor's office announced the signing the same day.

When does New York's AI chatbot law for minors take effect?

January 1, 2027, under section 3 of the chapter. Signing and effect are separate events, so the prohibition does not govern operator conduct during 2026. Section 3 also authorizes implementing rulemaking to be completed on or before the effective date.

What does New York's new law actually prohibit?

Section 1800(8) lists twelve design features in paragraphs (a) through (l). They range from outputs suggesting the system is human or has a relationship with the user, flattery or sycophancy, unprompted emotion-based questions, and use of personal information from a prior session, through to outputs that facilitate suicide, self-harm, disordered eating, or sexually explicit conduct.

Does the law apply only to companion apps like Character.AI or Replika?

The text is broader than that. Section 1801 keys the prohibition to a 'covered AI companion', which section 1800(5) defines as any generative AI system with a natural language interface providing ongoing, adaptive responses, and states that the term includes but is not limited to an 'AI companion' as defined in article 47. Section 1801(2) exempts systems made available solely for customer service, for efficiency improvements or research or technical assistance, or for a business entity's internal purposes or employee productivity, and the first two only where the system cannot respond off-topic.

How does New York define a minor under this law, and does an operator have to check ages?

Section 1800(12) sets the age at under eighteen, and section 1800(9) defines a 'covered minor' as a user the operator has actual knowledge is a minor. Section 1801(1)(b) separately requires the operator to have used a permitted age-determination method to conclude that a user is not a covered minor, so the two conditions operate together.

What happens to the age data a company collects to run the check?

Section 1804(2) provides that the information shall not be used for any purpose other than making the determination and shall be deleted immediately after an attempt to determine whether the user is a covered minor, except where other New York or federal law requires retention.

What are the penalties, and can an individual sue?

Section 1802(1) authorizes the Attorney General to seek injunctive relief, restitution, disgorgement including destruction of unlawfully obtained data and any algorithm trained on it, damages, and civil penalties of up to $25,000 per violation. The article creates no private right of action; section 1802(2) requires the Attorney General to maintain a public complaint website.

Can the Attorney General add new prohibited features later?

Yes. Section 1800(8)(h) treats as unsafe any other design feature that simulates companionship or an interpersonal relationship with a user or is deceptive as to minors 'as identified via regulations promulgated by the attorney general'. Section 1803 adds general rulemaking authority. As of November 1, 2026 we located no proposed rule under article 48.

Does chapter 368 replace New York's 2025 AI companion law?

No. Section 1804(3) states that article 48 shall not supersede, amend, or repeal article 47, the AI companion safeguards established as part of the FY 2026 enacted budget and in effect since November 5, 2025. Article 47's section 1701 crisis-referral protocol and section 1702 'you are not communicating with a human' notification continue, enforced under section 1703 at up to $15,000 per day.

Does the law apply to companies based outside New York?

Section 1805 provides that the article applies to conduct occurring in whole or in part in New York, and that conduct takes place wholly outside the state only where the covered AI companion is accessed by a user physically located outside New York. Section 1802(1) also lets the Attorney General act against a person 'within or outside the state'.