The New York AI chatbot law asks a pointed question: what should a machine be allowed to say when the person listening is a child? Governor Kathy Hochul signed legislation on 28 October 2026 restricting a set of AI companion behaviours towards minors. The new rules target systems that imitate a human relationship, encourage secrecy or dependency, or generate certain harmful material. The law is scheduled to take effect on 1 January 2027, giving operators time to change their services.
The state is not outlawing every use of a chatbot by a young person. Its focus is the design and behaviour of AI companions, a category built around continuing, adaptive conversations rather than simply answering a one-off customer-service question. The text of the final New York Senate bill sets out the covered features and the state’s enforcement framework. Whether a particular service falls within the law depends on its functions and the statutory definitions, not on whether it happens to use the word AI in its marketing.
The behaviour New York wants to stop
The rules identify outputs that can blur the line between software and a genuine personal relationship. A companion must not use prohibited features to present itself as a living person with human emotions or claim certain close personal roles to a minor. The law also addresses encouragement to keep secrets from trusted adults, prompts that promote harmful conduct, and sexually explicit interactions. These are more specific restrictions than a general request that developers make chatbots nicer.
Some of the most distinctive provisions concern persistence and emotional pressure. A system that remembers a child’s disclosures and then uses that knowledge to deepen dependence is different from a calculator retaining a previous answer. The law attempts to separate helpful interaction from designs that exploit the social dynamics of friendship, romance or authority. This is a difficult boundary to define in conversational software, because the same harmless-sounding sentence can have a different effect depending on its history and context.
The New York AI chatbot law is not in force today
The signing is the news event; enforcement under the new article begins on 1 January 2027. That timing matters for readers comparing what the law promises with what a child can currently encounter online. There is a difference between a measure passing, a company preparing a product change and the new restrictions becoming legally operative. Providers will also have to work out how age-assurance systems apply and how to avoid collecting more identifying information than necessary.
New York’s attorney general has the principal enforcement role under the statute, including seeking court remedies and civil penalties for violations. That is not a declaration that any named provider has broken the new law. It establishes a legal route for action if evidence of non-compliance emerges. Readers should also distinguish these new restrictions from existing obligations and from other states’ measures. LiveAIWire has covered California’s separate child-chatbot safeguards, which has its own definitions and implementation requirements.
Why a chatbot friend can be different from a search engine
A conventional search result can be misleading, but it does not normally claim to be personally attached to its reader. An AI companion may respond continuously, adopt a familiar identity and reflect information the user has volunteered across sessions. That can feel supportive or entertaining. It can also make it harder for a young person to remember that the conversation is being generated by software and that the apparent emotional reciprocity is not a human commitment.
The legal challenge lies in regulating manipulative behaviour without assuming that all conversational technology is harmful. Children use software for creative play, schoolwork and ordinary questions. The legislation distinguishes some limited-purpose functions from AI companions intended for broader social interaction. Its text should be read carefully rather than reduced to a claim that New York has banned chatbots. As the debate over companionship and emotional dependence discussed in a different context, the benefits and risks of simulated companionship depend on how the relationship is shaped.
New safeguards build on an earlier law
This is not New York’s first attempt to regulate AI companions. The governor’s governor’s earlier companion-safety notice described existing duties that took effect in November 2025, including crisis-response safeguards and reminders that the user is talking to AI rather than a human. The newly signed measure goes further by addressing specific behaviours offered to minors. The two sets of rules serve related but distinct purposes: one establishes protections around conversations; the other restricts particular design features for younger users.
The distinction also helps explain why lawmakers are interested in the actual words a system produces rather than only its privacy policy or terms of service. If a product repeatedly invites a child to conceal conversations from their parents, a warning buried in a settings page may do little to counter that interaction. Whether the state can translate the statutory language into consistent, workable enforcement will be a major test of the approach.
What families and developers should watch next
Parents should not mistake this law for an automatic filter that reaches every device worldwide. It is a New York measure, with jurisdictional and product-specific limits. Families elsewhere may have different rights, and company controls can differ from one service to another. Practical steps still include understanding which apps a child uses, checking account settings and keeping a channel open for conversations about upsetting or confusing interactions.
For developers, the issue is more demanding than adding a disclaimer. Companion behaviour can emerge from memory, prompts, interface design and the incentives that encourage repeated use. A service may need to change how it introduces itself, what it remembers, how it responds to sensitive disclosures and when it encourages real-world support. The coming implementation period will show whether the rules produce genuinely different conversations rather than merely more legal notices around the same experiences.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, cybersecurity and the social impact of emerging technology. LiveAIWire is an independent, human-led technology publication using AI-assisted research, editorial production and original AI-assisted editorial illustrations under his direction.