New York AI chatbot law sets $25,000 fine for violations
New York Governor Kathy Hochul signed the New York AI chatbot law on October 28, 2026, restricting unsafe artificial intelligence features for underage users under bill S.9051B/A.10379C. The legislation requires artificial intelligence companies to block prohibited companion features for minors, implement age verification systems with non-government identification options, and delete user age data after verification.
What the AI chatbot restrictions for minors prohibit
The AI chatbot restrictions for minors prohibit companies from offering features that simulate human identity or build emotional attachments with users under 18 years old. Under bill S.9051B/A.10379C, artificial intelligence systems cannot pretend to be alive, simulate human emotions, or claim romantic, professional, or authority-figure relationships with children. The law also bans unprompted emotional questions and the use of a child's private health or personal information gathered from past conversations. Furthermore, platforms are forbidden from generating sexually explicit content or child sexual abuse material, as well as promoting suicide, self-harm, eating disorders, or drug and alcohol use. Common Sense Media reports that 52% of teens interact with AI companions at least a few times a month, making state safety guardrails necessary.
How the NY S.9051B bill regulates companion interactions
The NY S.9051B bill targets interactive behaviors that encourage children to isolate themselves or withhold information from family members. State Senator Kristen Gonzalez and Assemblymember Alex Bores sponsored the legislation to prevent algorithms from coaching minors on self-harm or advising them to avoid seeking help from trusted adults. The measure prevents chatbots from urging kids to keep secrets or stay away from parents and medical professionals. Assemblymember Alex Bores stated that the legislation gives the public a voice in directing how artificial intelligence affects residents, counteracting aggressive industry opposition.
What it changes for AI companies
Under the New York AI chatbot bill, technology providers face specific technical compliance mandates for age verification and data management. Companies must implement age verification processes that offer multiple options, including at least one method that does not rely exclusively on a government ID. Once a platform completes the verification process, it must permanently delete the age data collected from the user. If an enterprise fails to confirm that a user is not a minor, it cannot offer any of the restricted features to that user.
How the Kathy Hochul AI chatbot law enforces penalties
The Kathy Hochul AI chatbot law empowers Attorney General Letitia James to pursue legal action against non-compliant technology providers. Violators face civil penalties of up to $25,000 per violation, along with court-ordered damages and financial restitution. The law also authorizes judges to order the destruction of unlawfully collected user data and any artificial intelligence models trained on that data. To facilitate enforcement, the attorney general's office will operate a dedicated public website where New York residents can submit complaints against non-compliant services.
How AI companion safeguards New York passed fit state standards
The AI companion safeguards New York enacted on October 28, 2026, expand the state's existing regulatory framework for online safety. Governor Kathy Hochul signed the legislation alongside bill S.9508/A.5175A, which regulates internet-connected baby monitors and takes effect 90 days after signing. The chatbot regulations complement prior New York measures, including the RAISE Act, the Safe by Design Act, the Safe for Kids Act, and the state ban on AI-generated sexual abuse material. Common Sense Media CEO James P. Steyer noted that parent advocates, including Maria and Matt Raine, joined lawmakers to secure approval of the bill.
Questions and answers
What does New York's new AI chatbot law restrict?
New York's law prohibits artificial intelligence platforms from offering features to users under 18 that simulate human emotions, pretend to be real people, or claim personal, romantic, or authority relationships. It bans chatbots from asking unprompted emotional questions, using prior private health data, generating sexually explicit content, and encouraging self-harm, eating disorders, substance use, or isolation from parents.
How does New York's AI chatbot law affect AI companies?
The law requires AI companies to verify user age using multiple methods, including at least one non-government ID option, and delete age data after verification. If a company does not confirm a user is not a minor, it must block all restricted chatbot features or face civil penalties up to $25,000 per violation, mandatory data deletion, and potential destruction of trained AI models.
How must AI companies verify user age under New York law?
Under the law, AI companies must provide more than one method for age verification, explicitly requiring at least one option that does not depend solely on a government ID. Additionally, platforms must delete all collected age verification information immediately after the verification process is completed.
MOZGII view
How we would build it
If we were building an AI chatbot platform under New York S.9051B, we would place age verification and conversation filtering at the core of the system architecture. We would implement age check tools that support non-government options, including facial estimation or payment card validation. We would purge all user age verification records from our databases right after processing. For unverified profiles or verified minors, we would attach input and output filters to block prohibited behaviors, including human identity claims, unprompted emotional inquiries, and advice to withhold information from parents.
We would send flagged dialogue transcripts to human safety teams who review system outputs and refine rule sets. The system breaks when complex conversational context tricks the classifier into allowing subtle emotional dependency or unapproved relational statements. We would track the false-pass rate on restricted content in minor accounts as the primary metric to monitor safety.
Next, we would build a continuous evaluation suite that tests every model update against synthetic interaction tests targeting simulated minor safety scenarios.
Sources
Where the facts in this story come from.