New York’s AI Companion Law: What Parents Should Check Now
New York’s new AI companion restrictions start January 1, 2027. See who they cover, what parents can check now and why a chatbot warning is no safety guarantee.
What happened
New York Governor Kathy Hochul signed S9051B on October 28, 2026, adding restrictions on unsafe AI chatbot functions for minors. The Senate records the signing as Chapter 368. The new restrictions take effect January 1, 2027, according to §3 of the final text.
The law targets features such as presenting a bot as human, claiming a relationship with its user and encouraging secrecy or isolation. It is broader than a rule about romantic companion apps, but it is not a blanket ban on every chatbot. AP and Fortune reported the signing; the statutory text determines the precise scope.
For your family, signing and implementation are different moments. You can check what your teen uses today without assuming that a future legal duty has already changed the app. New York also has earlier companion-chatbot disclosure requirements; the January date is specific to these new restrictions.
What it means for parents
The new law places duties on operators, with coverage tied to users physically in New York. It defines minors as under 18. Its “covered AI companion” definition is broader than its relationship-focused “AI companion” definition: it includes conversational generative systems with ongoing adaptive responses. The name “assistant” or “companion” does not decide the question. Specific exceptions cover tightly limited service or work functions; simply calling a general chatbot educational does not establish an exception.
The table maps selected provisions of the final S9051B, §§1800–1805, to a family decision. These are proposed checks, not duties imposed on parents or tests of a company’s legal compliance.
| What the law addresses | What you can check now | What that cannot prove | |
|---|---|---|---|
| Where and when | New restrictions from January 1, 2027; access physically outside New York is excluded. | Where your teen is physically located while using the service, and the provider’s current local terms. | That an app changed on signing day or that this rule applies nationwide. |
| Humanity and relationships | Listed functions include human/emotional deception and claims of personal or professional relationships or authority. | How the service describes its characters and any concerning interaction your teen wants help with. | That an “AI” label makes every subsequent response appropriate. |
| Secrecy and isolation | Encouraging secrecy, isolation or rejection of professional help or appropriate adults is listed. | Agree who your teen will tell if a bot asks for secrecy or discourages human help. | That no troubling exchange occurred just because a parent has not seen one. |
| Personal information and memory | Listed functions include using personal or health/well-being information from an earlier session or obtained more than 12 hours earlier. | The provider’s memory, personalisation and data-retention explanations. | That every chat record must disappear after 12 hours. Restricted use and deletion are different. |
| Age and permission | Operators must determine that a user is not a minor before providing restricted functions, using the statutory age-assurance framework; no parental-consent exception permits those functions. | Actual service eligibility and any separate consent requirement. | That a store rating or a parent’s approval overrides the law or stricter provider terms. |
The practical conclusion is modest: a law can set a boundary for operators without giving you evidence that an individual conversation meets it. Nor does this law create general permission for all 13-year-olds to use every service. For the wider relationship question, see our guide to teens and AI companions.
What to do now
Start with a conversation and a small, visible check together. AAP’s parent guidance supports talking about chatbot use and the difference between a machine and a person. The following sequence is our practical application of that guidance, not a clinical assessment.
- Name the services and their purpose. Ask which standalone apps and built-in bots your teen uses, and what they get from them. The result is a short shared list, rather than an assumption based on one app icon. If a service name is unclear, identify it together before deciding which rules apply.
- Check eligibility at the source. Open each provider’s current official terms for your country: minimum age, consent and feature restrictions. Record the relevant page and date. If you cannot establish eligibility, keep it unresolved and ask the provider. An app-store content rating alone is not the answer.
- Review privacy together. Following UNICEF’s guidance, look at what information is collected, used and retained. Find the explanations for memory and deletion; do not assume switching one off removes everything already stored. If a control or policy is unclear, write down that gap rather than mark privacy “done”.
- Agree on a human route. As a family rule, decide that a bot claiming a romantic bond or asking for secrecy is a reason to pause that interaction and tell a trusted adult. Name that person and how to reach them. You do not need to provoke harmful responses or recreate intimate content to make this agreement.
For this ordinary check, begin with the service list, terms and conversation; searching an entire chat history is not required. This is not a rule against looking together at a specific exchange when help is needed. Common Sense Media recommends avoiding social AI companions under 18; this is its own recommendation, separate from the statute and specific to that category.
A household example
A hypothetical 15-year-old in Albany uses a homework chatbot and a separate social bot. The parent has not checked either service’s terms. Applying the table gives the family a decision without pretending to have tested either provider.
- 1Check eligibility and privacy firstIdentify both services. In this example, the homework bot’s local age terms and retention explanation cannot be found. The parent asks its support team; eligibility and deletion remain unconfirmed. A homework label does not establish an exception.
- 2Act on the reported concernThe teen says the social bot claims a romantic relationship and asks for secrecy. Under their agreed family rule, they pause that interaction. The teen chooses a trusted aunt and calls her with the parent, without reproducing the exchange to test it.
- 3Keep the conclusion boundedThe homework service remains unresolved pending eligibility and privacy answers; the social interaction is paused and the aunt has been contacted. None of these decisions waits for January or treats the law as a safety certificate.
When human help comes first
Human help comes first if there is a crisis; a chatbot or a future regulation is not the support route. In the US, you can call or text 988 about someone you are worried about. For an immediate physical safety threat, call 911. Elsewhere, use local crisis and emergency services. A secrecy request alone is not a clinical diagnosis.
Dates and sources to keep straight
Use final S9051B, §3, for the January 1, 2027 date. Engadget’s October 28 report says six months; earlier bill versions used 180 days. The final signed version controls this article’s date, and the cause of the reporting discrepancy is not established.
This guide reflects sources checked October 30, 2026. Later amendments, implementing rules and a provider’s current local terms can change the relevant decision. Keep the immediate family check separate from a prediction about enforcement. Our broader AI risks guide covers other concerns beyond this New York event.
Frequently asked questions
When do New York’s new AI chatbot restrictions take effect?
The final signed S9051B sets January 1, 2027 as the effective date, while allowing preparation of necessary rules immediately. The October 28, 2026 signing did not itself switch on every new restriction. Older bill versions used 180 days, which is why checking the final version matters. New York also has earlier companion-chatbot safeguards; January is the start date for these new restrictions, rather than the first protection of any kind.
Does New York ban every chatbot for under-18s?
No. The new law restricts listed unsafe functions in covered services, with specified exceptions, and excludes access while the user is physically outside New York. Its definition reaches beyond apps marketed as companions, so calling a product an assistant does not settle coverage either. Do not treat this short summary as a verdict on a particular service. Provider terms, actual functionality and the statutory exceptions still matter.
Does parental consent make the restricted features permissible?
The new law does not provide a parental-consent exception to its prohibition on the listed unsafe functions. Do not import consent rules from a different law, such as rules for social-media feeds, into chatbot use. A provider can separately require adult consent under its own terms, but meeting that requirement does not override the statute. Check eligibility and the feature’s scope as separate questions rather than treating one parental approval as permission for everything.
Does a chatbot age rating mean my 13-year-old can use it?
An app-store rating alone does not establish that your child meets a service’s terms. Open the provider’s current official terms for your country and check its minimum age, adult-consent requirements and any feature-specific restrictions. New York’s law does not create a general permission for everyone aged 13 or older to use every chatbot. If the terms are unclear, keep eligibility unresolved and ask the provider before treating access as approved.
What should parents outside New York do now?
Do not assume this state law changes the service where your family is located. Check the rules and provider terms applicable to you. The practical family conversation still travels: identify the services your teen uses, discuss what a bot can and cannot be, review privacy together and agree whom to tell about troubling interactions. Those are suggested family actions, not a claim that New York’s requirements apply worldwide or that a checklist guarantees safety.