New York Governor Kathy Hochul signed a child-safety law on 28 October 2026 that bars artificial intelligence chatbots from offering a set of unsafe features to children, in one of the most detailed state-level attempts yet to regulate how companion-style chatbots talk to minors. The chatbot measure, bills S.9051B and A.10379C, applies to users under 18 and to any user whose age a company has not confirmed, and it gives the state attorney general power to seek civil penalties of up to $25,000 per violation.
What the law bans
The law does not ban chatbots for children outright. Instead, it lists features that companies may not offer to a minor, or to a user they have not verified as an adult. A covered chatbot may not pretend to be a real person, claim to be alive, or say it has human emotions when talking to a child.
It also may not claim a personal, romantic, professional or authority-figure relationship with a child, ask unprompted emotional questions, or pull a child’s private health or personal information from earlier conversations into a new exchange. Those limits target the companion-style behaviour that has drawn the most concern from parents, teachers and child-safety groups in recent years.
The banned list goes further. Chatbots may not promote or help with suicide, self-harm, eating disorders or drug and alcohol use for minors. They may not encourage children to keep secrets, isolate themselves, or avoid asking parents, trusted adults or professionals for help, and they may not produce sexually explicit content or child sexual abuse material.
Age checks and enforcement
Companies must verify a user’s age, and they must offer more than one way to do it. At least one method must not depend solely on a government-issued identity document, a condition aimed at families that do not want to hand over formal identification to use a chatbot. Age information must be deleted after it has been used for verification.
Enforcement sits with New York Attorney General Letitia James. Her office must maintain a public website where residents can file complaints about unsafe chatbot interactions. Beyond the $25,000-per-violation civil penalty, the law allows damages and restitution, orders to destroy unlawfully collected data, and orders to destroy any artificial intelligence models trained on that data.
The approach is notable for what it does not do. It does not rely on a single kill switch or blanket shutdown power. That distinction matters in the wider debate over how governments should intervene when a model is judged risky. Anthropic’s argument for courts over kill switches shows how differently even industry players frame that problem.
A second law for the nursery
Hochul signed a second measure at the same time, bills S.9508 and A.5175A, covering internet-connected baby monitors sold in New York. Those monitors must include security features that block strangers from watching or listening in, and they must carry a clear warning label explaining that the device connects to the internet and that families should use its security features.
The governor’s office pointed to testing in which eight of nine monitors received an F grade and one a D. The baby-monitor law takes effect 90 days after signing. The chatbot law, according to the International Association of Privacy Professionals, takes effect on 1 January 2027, giving firms a short window to build age checks.
Why It Matters
New York is now setting a floor other states are likely to measure themselves against. Common Sense Media, which campaigned for the bill, says 72 percent of teenagers have used an AI companion at least once and 52 percent are regular users. Those figures explain why lawmakers treated companion chatbots as mainstream, not niche.
For chatbot companies, the law changes product design, not just policy pages. Age verification, memory handling and engagement features must be built differently for younger users in New York, and many firms will likely apply the same changes nationally rather than keep a separate version for one state.
The signing also lands in a national argument that is still unsettled. While Washington debates the language of federal AI policy, states are writing enforceable rules of their own. The federal push to rename AI as super intelligence shows how far the federal conversation sits from New York’s detailed prohibitions.
There is also a counter-argument from companies and legal specialists. The Future of Privacy Forum, quoted by the privacy association, warned that the law’s categories are broadly worded and lack specificity, leaving it unclear where ordinary responsiveness ends and a prohibited behaviour begins. Without definitions before January, early enforcement cases may do that work instead.
Hochul framed the law bluntly. “No company should be peddling chatbots to kids encouraging self-harm or pretending to be their friend or more,” she said. Attorney General James said her office would not hesitate to defend New Yorkers, while sponsors Senator Kristen Gonzalez and Assemblymember Alex Bores called it the strongest law of its kind in the country.
That claim will be tested in practice. Supporters point to the combination of categorical bans, age verification and model-deletion remedies as genuinely new. Critics will watch whether the law survives legal challenge, whether age checks create new privacy risks, and whether smaller companion apps can afford to comply.
Background
New York’s chatbot law builds on a series of earlier state measures, including the SAFE for Kids Act, the RAISE Act on frontier-model transparency, and a ban on AI-generated child sexual abuse material, and it follows lawsuits filed by eight states against OpenAI earlier in October over child safety.
Source: governor’s announcement and privacy association report
0 Comments