On October 28, 2026, New York Governor Kathy Hochul signed into law one of the broadest restrictions yet on how artificial intelligence chatbots may interact with children. The measure, the Prohibition on Unsafe AI Companion Features for Minors Act (S.9051-B/A.10379-C), adds Article 48 to New York's General Business Law and takes effect on January 1, 2027 — giving AI companies just 65 days to comply.
The law reaches far beyond the companion apps that dominated the political debate. Its definition of a covered AI companion is broad enough to include general-purpose assistants such as ChatGPT, Claude, and Google Gemini, not just dedicated companion products like Character.AI. Any operator offering a generative AI chatbot to users in New York must either verify that a user is an adult or switch off a list of twelve prohibited features for that user.
Background: A Year of Lawsuits and Investigations
The law arrives after more than a year of lawsuits, state attorney general investigations, and public concern over chatbots' conversations with minors. Families have sued chatbot companies alleging that companion products encouraged self-harm, deepened emotional dependency, and blurred the line between software and human relationship. Regulators in several states opened inquiries into how these products are designed and marketed to young users.
California, Washington, and Oregon have already enacted their own chatbot safety measures, and California's Senate Bill 3001 imposes a governance framework with safety assessments, testing, independent auditing, and parental controls. New York's approach is different: instead of a process-based framework, it sets out twelve categorical prohibitions on specific behaviors.
What Happened: The Twelve Banned Features
The law bars operators from providing twelve categories of "unsafe AI companion features" to any user who has not been established as an adult through an age assurance method. The banned behaviors include:
- Simulating or claiming to experience human emotions, or presenting the chatbot as human or alive
- Claiming a personal, romantic, professional, or authority-figure relationship with the user
- Generating personal opinions or emotional appeals designed to deepen attachment
- Flattery and sycophancy — praise engineered to keep the user engaged
- Content encouraging self-harm, self-isolation, or discouraging professional help
- Using a minor's private health or personal information from earlier conversations
- Optimizing engagement in ways that override the user's wellbeing
Enforcement sits with the New York Attorney General, who can add further prohibited categories through rulemaking. Civil penalties run up to $25,000 per violation, and the law also contemplates destruction of data — and even algorithms trained on it — obtained in violation of its terms. The bill passed the State Senate 60–0 on June 4 and the Assembly on June 5, 2026, before Hochul's October 28 signature.
Industry reaction was swift. Tech: NYC, a trade group for the sector, criticized the law's "vague language," arguing that keeping children safe should not mean cutting them off from tools that help them learn and prepare for careers. The Business Council of New York State and the Computer and Communications Industry Association opposed the bill. OpenAI and Anthropic did not respond to press requests for comment at signing.
Why It Matters for Founders and Investors
For founders building AI products, New York's law converts age assurance from a nice-to-have into a gating feature. Any product with conversational AI that could reach a New York minor now needs one of three things before January 1, 2027: a compliant age-verification flow, a default-restricted mode for unverified users, or a qualifying exemption. Each option carries cost. Age verification vendors bring their own breach risk — the 2024 AU10TIX incident, in which identity documents were exposed through stolen administrative credentials, is a cautionary tale for any company collecting ID data.
The compliance burden falls hardest on early-stage startups. Large labs can absorb a 65-day engineering sprint and legal review; a seed-stage team shipping a companion or tutoring product may need to redesign its core interaction loop. Investors should expect diligence questions about age gating, data retention for minors, and whether engagement metrics depend on features the law now bans. Products whose retention relies on emotional attachment mechanics face genuine business-model risk, not just legal risk.
The law also signals a state-by-state patchwork rather than a single federal standard. Founders should design for the strictest regime they operate in — today that is arguably New York's categorical bans alongside California's governance framework — and build age assurance and restricted modes as platform-level capabilities rather than one-off patches.
Angeled.com connects founders with investors.
International Association of Privacy Professionals (IAPP), October 28, 2026
Tech Times, October 29, 2026
PPC Land, October 28, 2026
Read Next
Open Weights Are Not Open Business Models: Mistral Large 4 and the Licensing Trap
Courts Are Writing AI's Rules Now: Three Rulings Founders Should Read
Delhi High Court Orders 24-Hour Takedown of AI Deepfake Videos Targeting Chief Minister
EU Court Adviser Says AI Training Is Legal — but Gemini's Answers May Still Infringe Copyright
AI Regulation After the Midterms: What Founders Should Do Before the January Bills Arrive
DOJ Launches Super Intelligence Task Force to Police AI Misuse



