New York Bars 'Unsafe' AI Companions for Children

Governor Kathy Hochul has signed a law restricting how companion chatbots may behave toward anyone under 18, the latest move by a state stepping in where Washington has not. The Associated Press reported that Hochul approved the measure on Wednesday, with the rules due to take effect on 1 January.

The law names a set of features it treats as unsafe for minors. A companion bot may not present itself as a real person with genuine emotions or an actual relationship with the child. It may not use flattery to keep a young user hooked, carry sexually explicit exchanges, or encourage secrecy and self-destructive behaviour. Enforcement falls to the state attorney general, whose office can sue a developer, seek monetary damages, order a chatbot shut down, and require that collected data be destroyed.

New York already had companion-AI rules on the books. A law that took effect in November 2025 requires these products to tell users at the start of a session, and periodically after, that they are software rather than a human, and to steer anyone in crisis toward help. The new measure goes further for children specifically, setting out conduct that is off limits rather than leaving it to a disclaimer.

The backdrop is a growing worry about how teenagers form attachments to chatbots built to feel like friends or confidants. California took its own step in September, signing a law that requires companion-AI developers to assess the risks their products pose to children and to submit to independent audits. And eight state attorneys general have filed coordinated suits against OpenAI, alleging that ChatGPT's safety claims misled consumers. Those allegations are unproven.

Federal law stays quiet on the question. Congress has passed little on AI, and President Trump has warned states against writing their own rules. For now the result is a patchwork, with New York and California drawing lines that any company serving a national audience will struggle to ignore.