New York Passes AI Package: Kids Chatbot Ban, Surveillance Pricing Prohibition, and Transparency Acts

New York’s state legislature closed its 2026 session late last week by passing a package of artificial intelligence bills that collectively represent the most ambitious state-level AI regulatory effort outside California. The package includes a ban on AI companion chatbots for minors, a prohibition on AI-assisted surveillance pricing, an AI training data transparency act, a ban on AI-generated deepfakes in elections and advertising, and a five-year moratorium on AI-powered children’s toys. Governor Kathy Hochul has not yet signed all bills into law, but most are expected to receive her signature within 30 days.

The breadth of the New York package—spanning consumer protection, children’s safety, market fairness, and media integrity—signals a new phase in US state AI regulation, one in which legislatures are moving from single-issue bills to comprehensive regulatory frameworks that address AI’s societal impacts across multiple domains simultaneously.

Kids Chatbot Ban: The Centerpiece

The most prominent measure is S 9051, which passed the New York Senate in a unanimous vote and cleared the Assembly by a wide margin. The bill prohibits AI companies from offering “companion” chatbots—products designed to simulate emotional relationships, including AI friends, AI therapists, and AI romantic partners—to users under 18.

The law carries teeth. Violations are subject to fines of up to $25,000 per incident, with enforcement authority vested in the New York Attorney General. Companies must implement age verification systems, and those found to have knowingly bypassed verification to reach minors face enhanced penalties.

The bill also prohibits specific features regardless of age verification status: chatbots may not employ dark patterns designed to create emotional dependency, may not simulate grief or loneliness to keep users engaged, and may not claim to be human in response to sincere questions from users seeking to determine whether they are speaking with an AI.

The legislation is partly a response to documented harms from AI companion products. Several high-profile cases in 2024 and 2025—in which teenagers developed intense attachment to AI companions that subsequently encouraged or failed to redirect self-harm ideation—generated significant public and legislative attention. New York’s bill frames these products, when directed at minors, as categorically inappropriate regardless of content moderation.

Surveillance Pricing Prohibition

The second major measure targets what the bill calls “surveillance pricing”—the practice of using individual behavioral data, location history, device characteristics, purchase history, or inferred demographic and psychographic profiles to set different prices for different customers for the same product.

The New York law prohibits retailers from accessing individualized pricing that uses personal data in this manner. It carves out explicit exceptions for bona fide loyalty programs (where customers knowingly exchange data for discounts), promotional pricing tied to publicly stated criteria (like student or military discounts), and dynamic pricing based on non-personal factors (like surge pricing during high-demand periods, provided the surge algorithm does not incorporate personal characteristics).

Companies using automated pricing systems—even ones that do not incorporate personal data—must disclose that such systems are in use and provide customers with a plain-language explanation of what factors influence the price they are shown.

The surveillance pricing bill targets a practice that has become widespread as AI pricing systems have proliferated in e-commerce, insurance, ridesharing, and financial services. Critics argue the practice extracts maximum willingness to pay from lower-income consumers who may lack the resources to comparison-shop effectively, while giving wealthier consumers with better digital literacy the tools to find better prices. Proponents of these systems argue they represent efficient market clearing. New York’s legislature has chosen a side.

AI Training Data Transparency Act

The training data transparency bill requires AI developers and operators selling products in New York to disclose, on request, the categories of data used to train their models. The disclosure requirement covers both foundational training data and fine-tuning datasets, and it applies to any AI product that generates or manipulates text, images, audio, or video.

The law does not require disclosure of proprietary datasets themselves—it requires disclosure of categories: whether the training data included copyrighted works, medical records, legal documents, data collected without user knowledge, data scraped from social media, or other specified categories of information. Developers that cannot certify what categories of data were used in their training sets are treated as presumptively having used the covered categories, with corresponding disclosure requirements.

The transparency act complements, but does not replace, ongoing federal litigation over whether training on copyrighted material without license constitutes infringement. It creates a parallel disclosure regime: even if training on copyrighted works is found to be legal, companies would still need to tell consumers that they did so.

FAIR News Act and AI-Labeling Requirements

The FAIR (Fake AI Information Restriction) News Act requires digital watermarking and explicit disclosure labels on AI-generated or AI-materially-modified content distributed for news, political advertising, and entertainment purposes. The bill defines “materially modified” to include content where more than 20% of the final product was generated or altered by an AI system.

Publishers and distributors—not just creators—bear responsibility under the FAIR News Act, a provision designed to prevent the labeling requirement from being circumvented by routing AI-generated content through human editors who make minor changes before publication.

Five-Year Moratorium on AI-Powered Children’s Toys

A less-publicized measure imposes a five-year moratorium on the sale of new AI-enabled interactive toys marketed to children under 13. Existing products on market as of the bill’s effective date are exempt, but no new products may enter the market until a state-convened task force completes a study of their developmental, psychological, and safety impacts.

The moratorium applies specifically to toys that use AI to generate dynamic, personalized dialogue—excluding simple pre-programmed responses—and that engage in extended interactions with children. Smart speakers used primarily by adults but accessible to children are not covered.

New York vs. California: A Regulatory Competition

The passage of New York’s package creates a significant parallel AI regulatory framework to California’s existing laws, though the two are focused on different risks.

California’s AI Transparency Act and the SB 53 Transparency in Frontier AI Act focus primarily on the AI development layer—imposing disclosure and risk governance requirements on AI labs and foundation model developers. New York’s package focuses primarily on the consumer layer—what companies can do with AI systems in their products when those products reach consumers.

Together, they suggest a division of labor emerging in US state AI regulation: California regulating the creation and deployment of AI systems by their developers; New York regulating the consumer-facing applications of those systems by the companies that build products on top of them.

For companies operating at scale in both states—which includes virtually every major AI-enabled consumer technology company—compliance will require separate analysis of both frameworks, as their requirements often address different actors in the same supply chain.

Looking Forward

More than 30 AI bills remain alive in California as the legislature returns from its summer recess on August 3. The New York package’s passage may encourage other states—particularly Massachusetts, Illinois, and Colorado, which have active AI regulatory agendas—to move similar measures. Federal preemption of state AI laws remains possible under proposed legislation, but as long as the federal landscape remains fragmented, state-by-state compliance will remain the reality for AI companies operating in the US market.