New York City

This measure is proposed and binds nobody yet.

Sponsored by Council Member Virginia Maloney, it would let a person sue an artificial intelligence provider, meaning any person that makes an artificial intelligence model available for commercial or public use, for harm occurring in the city that a third-party user caused by misusing or maliciously using the model, including by circumventing its ethical guardrails, content filters or safety restrictions, where the provider failed to implement reasonable safeguards against that misuse and the harm was foreseeable.

A court could award damages, including punitive damages, injunctive relief and other appropriate remedies, and the bill would take effect immediately.

This measure is proposed and binds nobody yet.

Sponsored by Council Member Julie Menin, the Speaker, it would make it unlawful to market, offer for sale, sell or deploy an artificial intelligence model unless a third-party validator (a person other than an affiliate of the model's developer, engaged by the developer) has validated it on topics including task performance, data provenance, disparate impact and safety, and the model includes a shut-down capability, meaning the technical capability for a human operator to make it stop functioning temporarily or permanently.

The validator would have to certify to the developer and to the Office of Cyber Command whether the model has been validated and disclose any financial or other interest it has in the model or its developer, and the director of the Office of Cyber Command would write rules on those submissions and on validator qualifications.

The civil penalty would be $25,000 per instance of marketing, offering for sale, selling or deploying a non-compliant model and $25,000 per instance of falsifying a validation, and up to $25,000 for a violation of any other provision of the subchapter, recoverable before the Office of Administrative Trials and Hearings or in a civil action by the corporation counsel, and the bill would take effect 180 days after it becomes law.

This measure is proposed and binds nobody yet.

Sponsored by Council Member Nantasha M. Williams, it would let an elected official or a candidate for nomination for election notify any owner, licensee or operator of a system that generates materially deceptive audio or visual media (a video, image or sound recording intentionally manipulated to depict speech or conduct by a candidate, some or all of which did not occur, and so realistic that a reasonable person would believe it did) of their refusal to authorize it to produce such media of them within 120 days of a covered election.

For a candidate who gives that notice, the entity would have to implement a method preventing users from generating such media of that person within 60 days of a covered election, and the method would meet the duty if it is consistent with industry standards, not overly burdensome on the system, cost-effective to implement and maintain and up to date.

An entity that fails to comply and permits the unauthorized creation or dissemination of such media would be guilty of a misdemeanor punishable by a fine of not more than $2,500 per depiction, and the official or candidate could seek injunctive relief.

The bill would not apply where the system's outputs are processed by a third party with no ownership or control over the underlying generative model, would not make an entity liable where it implemented a method and still could not prevent a depiction, for incidental or unforeseeable depictions, or where no notice was sent, and would take effect 120 days after it becomes law.

This measure is proposed and binds nobody yet.

Sponsored by Council Member Frank Morano, it would put duties on a chatbot provider, meaning any person creating, distributing or otherwise making available an artificial intelligence model that generates text, audio, image or video simulating conversation: notify each user that they are talking to a chatbot rather than a human, publish a data security program, give users access to their own retained chat logs, assess the chatbot for risks of harm to users every month, and not imply that its outputs come from or equal those of a licensed or certified professional.

It would also restrict the processing of personal data and chat logs: that of a user the provider knows or has reason to know is under 18 could never be used for training and could be processed otherwise only with a parent's or guardian's affirmative consent, and an adult's could be used for training only with that user's consent.

It would limit advertising use, profiling, sale and retention of chat logs, bar retaliation against users who refuse consent to training or access their own chat logs, and make a provider liable for any injury its chatbot causes a user, with chatbots treated as products for product liability actions.

A civil penalty of not more than $25,000 per violation would apply, the corporation counsel could bring an action to correct violations, a person alleging a violation could sue for compensatory, injunctive and declaratory relief, and the bill would take effect 180 days after it becomes law.

This measure is proposed and binds nobody yet.

Sponsored by Council Member Carl Wilson, it would bar any person from disseminating, or causing to be disseminated, in the city a commercial message promoting an artificial intelligence model, directly or indirectly, that does not disclose whether a third-party validator validated the model under the proposed subchapter 27 of chapter 5, or that includes a materially false or misleading statement about a substantial risk the model poses, any measure taken or to be taken to manage that risk, or whether the model was validated.

That subchapter is the one Council bill Int 1100-2026 would add. Each violation would carry a civil penalty of up to $25,000, with a separate penalty under section 20-703 where a deceptive advertisement also violates section 20-700, and the corporation counsel could sue to recover penalties and obtain injunctive relief. The commissioner would write rules on the form of the disclosure, and the bill would take effect 180 days after it becomes law.