New Jersey Bill Classifies AI Chatbots as “Products” Subject to Strict Liability

A recently introduced New Jersey Assembly Bill seeks to substantially change the liability landscape for companies offering artificial intelligence chatbots. If enacted as introduced, Assembly Bill A5607 would expressly classify chatbots as “products” for product liability actions and hold providers liable for injuries caused by their use even when the provider exercised reasonable care or had no direct contractual relationship with the user. It would also impose substantial restrictions on advertising, profiling, model training, and other uses of chatbot data.

Introduced on October 5, 2026, A5607 goes far beyond the companion-chatbot laws already in effect in California and New York, which principally establish disclosure and safety requirements for systems designed to simulate human relationships.  The proposed New Jersey statute would reach a broader range of conversational software, potentially including customer-service and general-purpose business chatbots, while adding a distinct strict-liability framework.

Broad Coverage Beyond Companion Chatbots

A5607 defines an “artificial intelligence chatbot” as a software application, web interface, or computer program that simulates human conversation and interaction through text, images, video, or audio. It does not limit coverage to emotionally engaging systems or applications marketed as virtual friends, romantic partners, or mental-health companions.

Its definition of “chatbot provider” is similarly expansive. It includes individuals and business entities—and specified members, affiliates, subsidiaries, and beneficial owners—that create, distribute, or otherwise make a chatbot available. The current version of the bill contains no express exclusion for customer-service, productivity, research, or internal business applications.

Consequently, the proposal could affect both companies developing chatbot applications and businesses deploying third-party chatbots through their websites, applications, or customer-support channels. Whether a particular upstream model supplier qualifies would depend on its role in creating, distributing, or making the chatbot available; the bill should not be read as automatically covering every supplier in an AI technology stack. Its precise territorial reach also remains an issue requiring clarification.

A Brief Primer on Products Liability Cases

Product liability cases arise when a person is injured as a result of a defective product.  In such cases, both the manufacturer and the seller of the product can be held liable for injuries caused by the product even if they acted carefully and did not know about the defect.  This is strict liability, in that the focus is on the product’s safety—not whether the company was careless or negligent.

To prevail in a products liability action, the injured person typically must prove that: (1) the product had a manufacturing defect, an unsafe design, or inadequate warnings; (2) the defect existed when the product left the defendant’s control; and (3) the defect caused the person’s injury.

For example, if a defective battery causes a fire and injures someone, the manufacturer may be liable even if it followed reasonable quality-control procedures. But strict liability is not automatic liability for every injury involving a product—the claimant must still establish a defect and a causal connection to the harm, and applicable defenses vary by jurisdiction.

A Statutory Path to Strict Product Liability

Section 4 of Assembly Bill A5607 contains the proposal’s central liability provisions:

1.   A Chatbot is to be considered a “product” for purposes of a product liability action:  This removes a threshold argument that conversational software should be treated as a service rather than a product. From a products liability standpoint, a chatbot would be no different than a hot water heater or a child’s car seat.
‍

2.   A provider has a duty to ensure that its chatbot does not injure users:  This is a legal responsibility requiring developers and deployers to protect others from harm.
‍

3.   A provider is liable for any injury from use of its chatbot:  Liability would be imposed even if the provider exercised all reasonable care in designing and deploying the chatbot.  That does not mean every inaccurate answer would necessarily generate a successful lawsuit, as the plaintiff must prove their injury was caused by the chatbot.
‍

4.   Liability applies to developers and deployers: Just like in any other products liability action, both the developer that made the defective chatbot and the business that provided it to the user can be held responsible for injuries caused by the defect, even if neither knew about the defect or acted carelessly.

Implications for Developers and Businesses Deploying Chatbots

As many people have already learned to their detriment, there are multiple ways a chatbot might cause injury to a user, including:

-       Providing dangerous medical advice, such as recommending that a user stop prescribed medication without consulting a doctor.
‍

-       Encouraging self-harm or reinforcing suicidal thoughts instead of directing a vulnerable user toward appropriate help.
‍

-       Giving misleading financial guidance that causes a user to make an unsuitable investment or lose money.
‍

-       Exposing sensitive personal information shared in a conversation, potentially leading to identity theft, fraud, or embarrassment.
‍

-       Fostering unhealthy emotional dependence, particularly in children or vulnerable users who mistake simulated companionship for a genuine human relationship.

If enacted, A5607 would require companies to evaluate chatbot risk as more than a matter of accurate disclosures and responsible development. The proposed framework would make product architecture, data practices, distribution arrangements, and injury prevention relevant to multiple potential claims.

A5607 remains a proposal, but its significance is clear: it would move New Jersey beyond regulating how companion chatbots interact with users and toward imposing product-safety accountability across a much wider conversational-software market. The most consequential questions are whether its broad coverage survives amendment and how its injury standard would interact with existing product liability requirements and defenses.