When the New Jersey Appellate Division decided Trenton Renewable Power, LLC v. Denali Water Solutions, LLC, 470 N.J. Super. 218 (App. Div. 2022), it resolved what looked, on its face, like a fairly narrow discovery dispute: how much of a non-party engineering firm’s files a litigant could demand in a contract case about an anaerobic biodigester facility. Four years later, that ruling has become one of the more important tools available to any non-party facing a subpoena in New Jersey, especially given the dramatic increase in the use of generative AI in corporate workflows.

The Problem Faced by the Court

Our client, Symbiont Science, Engineering and Construction, designed and built the facility at the center of the underlying dispute between Trenton Renewable Power and Denali Water Solutions. Symbiont was not a party to that lawsuit. It was, however, served with a subpoena covering 17 deposition topics and 13 categories of documents. Symbiont’s efforts to comply with that subpoena revealed an estimated 5.6 million pages of potentially responsive material and tens of thousands of dollars in processing and storage costs, with no end in sight.

The trial court treated the subpoena the same way it would have treated any document request between the actual litigants: broadly, and in favor of production. The Appellate Division reversed, holding that trial courts cannot simply import the liberal, pro-discovery standard that governs disputes between parties when considering discovery directed to non-parties. Instead, a stranger to the litigation is entitled to real, independent scrutiny of whether the same information is available via other less burdensome means, and whether the cost of production is proportionate to what the requesting party actually needs.

The Decision’s Impact Before the Advent of AI

Even before generative AI completely changed how companies create and store information, Trenton Renewable gave non-parties three concrete arguments that had been largely absent from published New Jersey case law:

• The “go ask a party first” argument. If the information sought from the non-party is burdensome for it to produce, and that same information likely exists in the files of an actual litigant, the requesting party should be pursuing it from the party first.
• The volume-as-burden argument. Sheer data volume, standing alone, can establish undue burden, even absent a privilege or confidentiality objection. Identifying, reviewing, and storing large volumes of data is costly and time-consuming, and the attendant expense and inconvenience must be considered by a court in the non-party context.
• The case-specific proportionality argument. Rule 4:10-2(g)’s proportionality factors require a genuine, case-specific weighing of cost against benefit, particularly where the target of discovery has no stake in the litigation’s outcome.

Why Trenton Renewable Matters More Today

Every one of those pressure points has intensified since 2022 because the volume and complexity of electronically stored information that companies generate has grown dramatically, and generative AI is a major contributor.

Courts are now confirming that content created by AI such as prompts, model outputs, and activity logs from tools like ChatGPT or Claude can be discoverable. Traditional ESI principles apply with equal force in the AI context. That means the categories of potentially discoverable material sitting inside any company’s servers have multiplied almost overnight. A company that gets pulled into someone else’s lawsuit today faces not just the discovery of its email archives and shared drives — it may also have to produce prompt histories, draft outputs never meant to see daylight, and activity logs generated by AI tools adopted for entirely unrelated business purposes.

Recent decisions around the country show that courts are confronting these issues:

• In United States v. Heppner, 820 F. Supp. 3d 292 (S.D.N.Y. 2026), a criminal securities fraud case, the court held that a defendant’s written exchanges with a generative-AI platform were not protected by the attorney-client privilege or the work-product doctrine, and that a user retains no substantial privacy interest in conversations he voluntarily hands to a publicly accessible AI platform that keeps them in the ordinary course of business. AI’s novelty, the court explained, “does not mean that its use is not subject to longstanding legal principles.” at 299.
• In Villanueva v. Las Vegas Metropolitan Police Department, No. 2:24-cv-00125, 2026 WL 2942348 (D. Nev. Sept. 29, 2026), an expert had used ChatGPT to help research and draft his report, and the court held that the AI-generated case summaries he consulted were “facts or data” he considered and had to be produced like any other source material. It shielded only the portions that revealed the expert’s own thinking — his prompts, a deposition outline, and draft-report text — as work product.
• In In re OpenAI, Inc., Copyright Infringement Litigation, 800 F. Supp. 3d 602 (S.D.N.Y. 2025), the court viewed a motion to compel the content of a party’s AI chat logs, including both the prompts entered and the outputs returned, through the ordinary relevance and proportionality tests that govern any electronically stored information. It denied production there as disproportionate, conservatively estimating that a mere privilege review of the 80,000+ log entries alone would take nearly 11 weeks and cost upwards of $1 million.

Moreover, as companies increasingly turn to outside AI platforms to draft, summarize, and analyze information, courts will need to consider whether those platforms function as neutral tools or as third parties with discoverable records of their own. If courts start treating AI vendors as potential targets for discovery, Trenton Renewable’s approach of favoring non-party burden, availability from a party, and proportionality will remain squarely on point for a category of “non-party” that didn’t meaningfully exist when the case was decided.

Discovering Non-Party Data in the AI Age

Trenton Renewable was decided in a world of email archives and shared drives. Now its holding must be applied to the world of AI-generated materials and related prompt logs. However, its core tenet still applies: Non-parties are not parties, and the burden of discovery imposed on non-parties demands its own, more scrutinizing, analysis. The distinction becomes ever more crucial as data volumes and the sources of discoverable information expand. It can be the only thing standing between a company with no stake in the litigation’s outcome and subpoena compliance that would cripple its operations and cost far more than the case is worth to anyone but the parties litigating it.

Daniel J. Cohen is a partner and Daniel C. Stark is an associate at A.Y. Strauss in its Hackensack office. They represented the winning appellant, Symbiont Science, Engineering and Construction, Inc., in Trenton Renewable Power, LLC v. Denali Water Solutions, LLC, 470 N.J. Super. 218 (App. Div. 2022).

About A.Y. Strauss

A.Y. Strauss provides legal counsel for commercial transactions and litigation in the areas of real estate, construction, employment, cannabis, family law, environmental, and corporate bankruptcy. With the intellectual depth of a large firm and the personalized touch of a boutique, the firm’s lawyers handle a wide variety of matters for emerging businesses, institutional clients, and high-net-worth individuals.