New Jersey AI Chatbot Liability Bill: A5607 Would Make Chatbot Makers Strictly Liable for Injuries

The New Jersey AI chatbot liability bill, Assembly Bill A5607, would declare every AI chatbot a "product" under state product liability law and make its provider liable for a user's injuries even if the company exercised all reasonable care. It also gives harmed users a private right of action in Superior Court worth the greater of actual damages or $5,000 per violation plus attorney fees, and it cleared the Assembly Public Safety and Preparedness Committee on Oct. 19, 2026, which is why defense lawyers quoted this week by the New Jersey Law Journal and the ABA Journal warn it could "open the floodgates."

No other state has gone this far. California and New York regulate how companion chatbots behave and, in California's case, let injured users sue for statutory damages. A5607 instead changes the liability rule itself. It removes the question most chatbot injury suits now fight over, whether a large language model is a product or a service, and replaces fault with strict liability. For the plaintiffs' bar it is the most consequential AI bill pending in any legislature. For chatbot companies, insurers and the lawyers who write consumer terms of service, it raises a second question nobody has answered yet: how much of this exposure ends up in arbitration rather than in a New Jersey courtroom.

What New Jersey bill A5607 would do

A5607 was introduced on Oct. 5, 2026 in the 222nd Legislature. Its primary sponsors are Assemblyman Cody D. Miller (D-4), Assemblywoman Marisa Sweeney and Assemblyman Joe Danielsen (D-17). Its official synopsis is modest: it "establishes content and data privacy requirements for chatbot providers." The liability provisions sit inside that privacy frame, and they are what has drawn national attention.

The bill is modeled on the People-First Chatbot model bill written by the Electronic Privacy Information Center (EPIC), the Consumer Federation of America and Fairplay. It has four working parts.

• Product status. A chatbot "shall be considered a product for the purposes of a product liability action."
• A duty and a liability rule. A chatbot provider has a duty to ensure that use of its chatbot does not cause injury to a user. The provider is liable for any injury to a user from use of its chatbot "even if the provider exercised all reasonable care in the design and distribution of the chatbot."
• Privacy limits. Providers could not process personal data other than the user's input data to generate outputs, unless the user expressly asks. According to the sponsors and bill summaries, providers could not sell conversations, use chat history to target ads, or train models on conversations without explicit consent. Chat logs could be kept for no more than two years unless other law requires longer.
• Ongoing safety duties. Providers would have to assess their chatbots for risks of harm every month and mitigate those risks under rules adopted by the Attorney General. Bill summaries also describe added protections for minors, including a bar on training with a minor's logs when the provider knows or should know the user is under 18, and parental consent before processing data of children under 13.

"Chatbot provider" is defined broadly. Bill summaries quote it as covering any individual, partnership, association, firm or business entity, and any member, affiliate, subsidiary or beneficial owner, that "creates, distributes, or otherwise makes available a chatbot." That reaches the model developer, the app that wraps the model, and potentially the corporate parent.

Would New Jersey make AI chatbot companies strictly liable for injuries?

Yes. If enacted as introduced, A5607 would make chatbot providers strictly liable for injuries their chatbots cause to users. The text says the provider is liable even if it exercised all reasonable care, so a plaintiff would not have to prove negligence.

That is a real departure from how New Jersey product cases work today. Under the New Jersey Product Liability Act (N.J.S.A. 2A:58C-1 et seq.), a plaintiff must prove the product was not reasonably fit, suitable or safe because of a manufacturing defect, a design defect or an inadequate warning. A Design Defect claim in New Jersey usually turns on risk-utility balancing, including whether a feasible safer alternative design existed. A Failure to Warn claim asks whether the warning was adequate. Both bring reasonableness back into what is nominally "strict" liability.

A5607's text skips that analysis. It does not ask whether the chatbot was defective. It asks whether the user was injured "from the use of" the chatbot. Read literally, that is closer to the absolute liability courts reserve for abnormally dangerous activities than to ordinary product liability. Expect the defense bar to argue that courts should read a defect requirement back into the bill through the Product Liability Act, and expect plaintiffs to point to the plain "all reasonable care" language. Causation stays in the case. A plaintiff would still have to show the injury came from use of the chatbot, and in self-harm and wrongful death cases that is often the hardest element.

Can you sue an AI chatbot company under the New Jersey bill?

Under A5607, a user harmed by a violation of the bill's core duties (Sections 2 and 3, according to bill summaries) could bring a civil action against the chatbot provider in the Superior Court of the county where the user lives. A prevailing user would recover:

• the greater of actual damages or $5,000 for each violation; and
• reasonable attorney fees and court costs.

The bill also gives the Attorney General separate enforcement power: a civil penalty of $5,000 for each violation and a civil action for injunctive relief against a provider that violates the bill or rules adopted under it.

Two features matter for plaintiffs' counsel. First, the venue clause points to the user's home county, which takes away the defendant's choice of forum, at least on paper. Second, fee-shifting plus a statutory damages floor turns low-dollar privacy violations, such as an unauthorized training use or retaining logs past two years, into claims a Contingency Fee practice can afford to bring. Scott Christie, a McCarter & English partner, put the risk plainly to Law.com: $5,000 in statutory damages per violation, no need to prove harm for the statutory claim, and a strict liability standard could push a single lawsuit into the hundreds of thousands or millions of dollars.

How much could a chatbot company owe under A5607?

The bill does not cap aggregate statutory damages, and it does not define what counts as a single "violation." That ambiguity drives the exposure math. The table below is illustrative only. It applies the $5,000 floor and is not a prediction of any award.

Scenario (illustrative)	How violations are counted	Statutory floor
One user, one unlawful training use of chat logs	1 violation	$5,000 plus fees
One user, 200 conversations used for training, counted per conversation	200 violations	$1,000,000 plus fees
1,000 New Jersey users, one violation each	1,000 violations	$5,000,000 plus fees
1,000 users, 50 counted violations each	50,000 violations	$250,000,000 plus fees
Personal injury claim (self-harm, medical harm)	Actual damages	Compensatory damages, plus any Punitive Damages under the Punitive Damages Act

Personal injury claims would rarely depend on the $5,000 floor, because actual damages in a wrongful death or serious-injury case are far higher. The floor matters most for the privacy-type violations that affect many users at once. That is where aggregation, whether a Class Action in court or Mass Arbitration under the provider's terms, decides what the bill is really worth.

Has the New Jersey AI chatbot liability bill passed?

No. As of Nov. 3, 2026, A5607 has cleared one Assembly committee and has not passed either house. Here is the record so far:

Date	Event
Oct. 5, 2026	A5607 introduced; referred to the Assembly Public Safety and Preparedness Committee
Oct. 19, 2026	Committee advances the bill and refers it to Assembly Appropriations; Assemblyman Antwan McClellan (R) votes no. Reports put the tally at 5-2 or 5-1
Oct. 20, 2026	EPIC says the bill is next set for the Appropriations Committee "on Thursday"
Oct. 22, 2026	Assembly Appropriations hearing listed on the legislative calendar
Oct. 30–Nov. 3, 2026	NJ Law Journal and ABA Journal coverage; defense bar warns of a litigation surge
Nov. 3, 2026	No floor vote recorded in the Assembly; no Senate companion identified in public trackers

The bill would still need Appropriations approval, a full Assembly vote, Senate committee and floor action, and the governor's signature. New Jersey's two-year session runs into January 2028, so the bill does not die at the end of 2026. That leaves room for amendments, and most bills this ambitious get them. Industry groups have already lined up against it. The New Jersey Chamber of Commerce told the committee that the state's comprehensive privacy law already covers some of the data concerns. The New Jersey Business and Industry Association argued that steep fines could hit smaller firms hardest. EPIC, the Consumer Federation of America and the Transparency Coalition testified in support.

A5607 also travels with other AI bills. A5090, the GAI Accountability Act, introduced May 14, 2026, would impose a $20,000 civil penalty per violation on generative AI owners, operators and developers that fail to prevent conduct that would be a crime if a person committed it. It offers an affirmative defense for "reasonable efforts" and relies on Attorney General enforcement. A5607 has no comparable reasonable-efforts defense, which is part of why the defense bar treats it as the more dangerous of the two.

Are AI chatbots products under product liability law?

Courts have not settled the question, and that is the gap A5607 would close by statute. Product liability law has traditionally treated software as a service rather than a product, and information and expression as outside it altogether.

The leading decision so far is Garcia v. Character Technologies, No. 6:24-cv-01903 (M.D. Fla.). In May 2025, U.S. District Judge Anne Conway let product liability claims over a 14-year-old's suicide proceed. She held that the Character.AI app could be treated as a product to the extent the claims arise from design defects "rather than ideas or expressions within the app." In the same 49-page order she declined, "at this stage," to treat the chatbot's output as protected speech, and she dismissed Alphabet. Garcia later settled. Since then, newer suits in California, including the coordinated OpenAI proceedings, have pressed the same design-defect theory.

A5607 would make that debate unnecessary in New Jersey, but only within the limits of state power. Two defenses sit above any state statute:

• Section 230 of the Communications Decency Act (47 U.S.C. § 230). Providers will argue some claims treat them as the publisher of third-party content. That argument is weaker when the harmful words come from the company's own model rather than another user, but it has not been resolved on appeal for generative AI.
• The First Amendment. A legislature can call a chatbot a product, but it cannot redefine speech. If courts eventually hold that model outputs are protected expression, strict liability for what a chatbot "says" will face constitutional limits that a product label does not avoid.

Federal preemption may also come from Congress. The bipartisan AI LEAD Act (S. 2937), introduced Sept. 29, 2025 by Senators Dick Durbin and Josh Hawley, would create a federal product liability cause of action for AI systems, with a four-year Statute of Limitations and added protections for minors. Unlike A5607, it is built on a duty of reasonable care in design and warnings, not strict liability.

How does A5607 compare with California SB 243 and New York's chatbot law?

A5607 is broader than either enacted state law, because it changes the liability standard for all chatbots rather than regulating particular companion-bot features.

Feature	New Jersey A5607 (pending)	California SB 243 (in force)	New York S9051B/A10379C (signed)	AI LEAD Act, S. 2937 (pending, federal)
Status	Passed one Assembly committee Oct. 19, 2026	Signed Oct. 13, 2025; core duties effective Jan. 1, 2026	Signed Oct. 28, 2026; effective date reported as June 1, 2027	Introduced Sept. 29, 2025; in Senate Judiciary
Scope	All chatbots and providers	Companion chatbots	AI companions interacting with users under 18	AI systems, developers and deployers
Chatbot declared a "product"	Yes	No	No	Yes, for federal claims
Liability standard	Strict: liable despite all reasonable care	Statutory duties (disclosure, self-harm protocols)	Prohibited features (12 categories)	Reasonable-care design and warning duties
Private right of action	Yes, Superior Court of user's county	Yes, for injury in fact	Not confirmed in reporting reviewed	Yes, federal court, class actions allowed
Damages	Greater of actual or $5,000 per violation, plus fees	Greater of actual or $1,000 per violation, plus fees	Not confirmed in reporting reviewed	Compensatory; four-year limitations period
Government enforcement	AG: $5,000 per violation, injunctions	Annual reporting to the state Office of Suicide Prevention	Not confirmed in reporting reviewed	U.S. Attorney General and state AGs

The per-violation floor is five times California's, and California's law has already drawn commentary that "injury in fact" will be heavily litigated. A5607's private action is tied to violations of the bill's duties, and the product designation applies across the board, so a New Jersey plaintiff would have both a statutory claim and a strict product liability claim in the same complaint.

Could arbitration clauses block lawsuits under the New Jersey chatbot bill?

Possibly, for many users. The bill's Superior Court venue clause does not override a valid Arbitration Clause in a chatbot provider's terms of service, and a state law that tried to would face FAA Preemption. Under AT&T Mobility v. Concepcion (2011), states cannot condition enforcement of consumer arbitration agreements on the availability of class procedures. Many consumer AI products are offered under terms that contain an arbitration agreement, often with a Class Action Waiver. Users who accepted those terms will be the first test of whether a New Jersey statutory claim stays in court.

Three exceptions limit how far arbitration reaches:

• Wrongful death claimants. Heirs and estates bringing their own statutory claims are often not bound by an agreement the decedent signed, depending on state law.
• Minors. Minors can generally disaffirm contracts, and courts have split on enforcing app terms against them. That matters because A5607's minor protections are among its most litigable parts.
• Formation defects. Browsewrap and weak clickwrap flows continue to lose motions to compel arbitration in federal court.

Where arbitration does apply, the $5,000 statutory floor changes the economics of Mass Arbitration. Each claimant's demand carries a guaranteed minimum value plus fee-shifting, and the respondent pays most of the provider's Filing Fee and arbitrator costs under AAA and JAMS consumer rules. A thousand New Jersey users with identical training-consent claims would present a $5 million statutory floor before fees, which is exactly the pressure that has pushed respondents toward batching protocols and global deals. Drafters at AI companies should expect A5607 claims to test their mass-filing provisions as soon as the bill takes effect, if it does.

What it means for plaintiffs' firms, defendants, drafters and funders

• Personal injury firms. A5607 would remove the product-versus-service fight that dominates early motion practice in chatbot injury suits and would drop the negligence element. Causation and damages would still require experts, especially in self-harm and medical-advice cases. New Jersey residency and use within the state will matter for intake, and the home-county venue clause favors local counsel.
• Chatbot providers and their insurers. Strict liability with no reasonable-efforts defense is close to uninsurable at current pricing for consumer-facing general-purpose models. Expect lobbying for a defect requirement, a safe harbor tied to Attorney General rules, a definition of "violation," and an aggregate cap. Monthly risk assessments would also create a discoverable paper trail in every case.
• Terms-of-service drafters. Arbitration agreements, delegation language, minors' consent flows and mass-filing provisions will carry more weight if the bill passes. Clauses drafted for $50 consumer disputes will look different against a $5,000 statutory floor per violation.
• Litigation funders. A statutory floor, fee-shifting and a clear liability rule make portfolio funding of chatbot claims easier to underwrite. New Jersey and federal disclosure debates over Third-Party Litigation Funding will follow the money.
• Mass tort practitioners. If similar bills spread, chatbot harm could become a coordinated Mass Tort docket. New Jersey already uses multicounty litigation for product cases, and a statute that sets the liability rule would make coordinated proceedings easier to justify.

This analysis describes the bill as introduced. Amendments in Appropriations or on the floor could change any provision discussed here.

FAQ

What is New Jersey bill A5607?

A5607 is a New Jersey Assembly bill introduced Oct. 5, 2026 that sets privacy and content rules for AI chatbot providers, declares chatbots "products" for product liability purposes, and lets harmed users sue in Superior Court for the greater of actual damages or $5,000 per violation.

Would A5607 make AI companies strictly liable?

Yes. As introduced, a chatbot provider would be liable for injuries to a user from use of its chatbot even if it exercised all reasonable care in design and distribution, so plaintiffs would not need to prove negligence.

Has the New Jersey AI chatbot liability bill become law?

No. It passed the Assembly Public Safety and Preparedness Committee on Oct. 19, 2026 and was referred to Assembly Appropriations. As of Nov. 3, 2026, it has not passed either house and has not been signed.

Who can sue under the New Jersey chatbot bill?

A user harmed by a violation of the bill's core provisions could sue the chatbot provider in the Superior Court of the county where the user lives. The Attorney General could separately seek $5,000 civil penalties and injunctions.

Does A5607 stop chatbot companies from forcing arbitration?

The introduced bill does not address arbitration, and the Federal Arbitration Act (FAA) would likely preempt any state rule barring enforcement of consumer arbitration agreements. Valid arbitration clauses could send many claims to individual arbitration, though wrongful death heirs, minors and users with defective click-through agreements may stay in court.

How is A5607 different from California's chatbot law?

California's SB 243 regulates companion chatbots and allows $1,000 per violation for injured users. A5607 covers all chatbots, sets a $5,000 floor, and adds strict product liability, which California's law does not.