← All posts
This post is written in English.
New York Governor Kathy Hochul signed a youth AI chatbot safety law on 28 October. It takes effect on 1 January 2027, which isn't much time for anyone building companion products.
What it requires
- Banned features for minors. Developers can't design chatbots that pretend to be human, encourage self-harm, or have inappropriate conversations with children, within 12 broad categories of unsafe features.
- Health data limits. Chatbots may not keep sensitive information about a minor's physical and mental health.
- Age checks. Providers must use multiple age assurance methods, at least one not relying on government ID, and delete age data after verification.
- Enforcement. Penalties reach $25,000 per violation, plus injunctive relief, with an attorney general complaint portal.
The open question
The law aims at design. But where does normal chatbot behaviour end? Future of Privacy Forum analyst Justine Gluck told IAPP "it's unclear where ordinary conversational responsiveness ends and prohibited behavior begins."
A chatbot that says "I'm glad you told me" is responsive. Is it also pretending to be a friend? That's the kind of question that may end up with lawyers.
My read
Regulating features is more interesting than regulating content, because it targets what keeps a child talking and not just what gets said. Age checks are the awkward part: to protect minors, you have to check everyone, which creates its own privacy risk. Requiring a non-ID option and deleting the data helps. Whether it holds up in practice is what I'd watch.
Twelve states had passed some form of consumer chatbot law as of July, according to the same report, so this isn't just a New York issue.
