The use and development of Artificial Intelligence ("AI") has accelerated exponentially in recent years, as companies rush to solve historical industry inefficiencies with big data solutions, and shows no sign of slowing down in the coming years. In the fourth quarter of 2024, more than half of the global venture capital was distributed to companies with AI-based technology. Alex IrvinHunt, AI Dominates Venture Capital Funding in 2024, fDiIntelligence (Jan. 8, 2025).
But whereas AI is increasingly a part of the business landscape, courts are just beginning to understand the technology's particularities. Arbitration associations such as the Silicon Valley Arbitration & Mediation Center ("SVAMC"), the American Arbitration Association ("AAA"), and JAMS, on the other hand, have more quickly adapted to meet technological and practical needs to accommodate AI by incorporating rules and guidelines intended to govern AI-related disputes and set protocols for the use of AI in proceedings generally.
As we see exponential growth in the development of AI, it is likely that we will see exponential—or close to it—growth in AI-related litigations and disputes. What only time will tell is whether the courts and dispute resolution bodies are ready for the potential influx of AI disputes and the novel issues and considerations that come along with them.
Silicon Valley Arbitration & Mediation Center
SVAMC released Guidelines on the Use of Artificial Intelligence in Arbitration on April 30, 2024. SVAMC Guidelines on the Use of Artificial Intelligence in Arbitration, SVAMC, 1(2024). The document is intended to guide the use of AI tools in arbitration proceedings, requiring participants who choose to use AI tools in arbitration preparation and proceedings to "make reasonable efforts to understand each AI tool's relevant limitations, biases, and risks and, to the extent possible, mitigate them." Id. at 9.
Notably, the SVAMC does not require parties disclose the use of AI tools, but prohibits the use of such tools to falsify evidence or mislead opposing parties or arbitrators. Id. at 11. The Guidelines do not explicitly prohibit the use of an AI tool to generate information outside of the record, but requires parties make appropriate disclosures and requires the arbitrator to "independently and critically" evaluate AI-derived information. Id. at 20.
American Arbitration Association
AAA released a document titled Principles Supporting the Use of AI in Alternative Dispute Resolution in November 2023. The paper is brief, less than two pages, and intended to reinforce the responsibilities of legal professionals in arbitration proceedings. Specifically, they warn against the assumption that "a technology can obviate a professional obligation." Id. at 1. The Principles require arbitrators and parties to exercise independent judgment over AI outputs, as arbitrators are ultimately responsible for their work product. Id. at 2. The publication does not speak to disclosure requirements.
JAMS
On April 23, 2024, JAMS released the arbitration rules to govern disputes involving Artificial Intelligence (JAMS Artificial Intelligence Disputes Clause and Rules, JAMS (June, 14, 2024), ("JAMS AI Rules")) announcing that "JAMS aims to equip legal professionals and parties engaged in dispute resolution with clear guidelines and procedures that address the unique challenges presented by AI, such as questions of liability, algorithmic transparency, and ethical considerations" (JAMS Announces New Artificial Intelligence Disputes Clause and Rules, JAMS (Apr. 23, 2024). The order is directed at disputes involving AI technology, rather than using AI in the preparation of materials used in the litigation of disputes. Id.
Definition of AI. Notably, JAMS defines "AI" broadly, as "machine-based system capable of completing tasks that would otherwise require cognition." JAMS AI Rules at Rule 1(e). This inclusive definition potentially covers a broad range of technologies, as tasks requiring human cognition can include "all forms of knowing and awareness, such as perceiving, conceiving, remembering, reasoning, judging, imagining, and problem solving." Cognition and the Brain, American Psychological Association (last visited March 25, 2025). This expansive definition of AI could allow for the JAMS AI Rules to apply to any technology with basic memory functions. Nearly all computers utilize both randomaccess memory (RAM) and hard-drives for the storage of data, which are commonly analogized to long- and short-term human memory.
Alterations from the standard JAMS Comprehensive Arbitration Rules include:
- The JAMS AI Rules closely mirror the accelerated proceedings that may be optionally applied per the JAMS Comprehensive Arbitration Rules & Procedures;
- The guidelines created procedures covering the production and expert review of AI product technical information; and
- JAMS created a protective order specifically for the use in AI technology disputes that provides increased levels of confidentiality designations as default.
Id. at Rule 16.1.
Production of AI Systems. JAMS's rules outline a specific procedure for the production of "any AI systems or related materials, including, but not limited to, hardware, software, models and training data," which must be made available to "one or more expert(s) in a secure environment" for purposes of litigating the dispute. Id. Additionally, at the parties' request, an arbitrator will select one or more experts to inspect the AI systems and produce related materials. Experts are selected from a list of experts maintained by JAMS, and the output of the expert is limited to a written report addressing questions posed by the arbitrator. Id.
Expedited Procedures. JAMS incorporated its pre-existing expedited dispute process, which is optional in disputes generally, into the Rules. Id. Absent other prescribed rules, disputes or claims administered by JAMS that are AI-related are governed by these expedited procedures. Id.
Protective Order. JAMS also adopted an AI Disputes Protective Order that applies to disputes covered by the Artificial Intelligence Disputes Clause. Id. The protective order creates both "confidential" and "highly confidential" designations for non-publicly available materials. Protective Order Pursuant to the JAMS Artificial Intelligence Disputes Rules, JAMS, (June, 14, 2024).
Courts
By litigating in courts, rather than an Alternative Dispute Resolution ("ADR") forum, companies with AI technology at the center of a dispute can face extended timelines for resolution. In 2024, the median time of a federal civil trial from filing to completion was over 33 months. Table T3- U.S. District Courts – Trials Statistical Tables for the Federal Judiciary (Dec. 31, 2024), United States Courts, (Dec. 31, 2024). While litigation timelines are largely agnostic to a company's product, discovery procedures vary by court and can have additional repercussions for AI-focused technologies.
Production of AI Systems. The extent to which AI core technology is discoverable is determined by the Federal Rules of Civil Procedure and local court rules. We are only just beginning to see how courts will approach the scope of discovery with respect to artificial intelligence's immense training datasets and iterative algorithms.
A recent case highlights the extent to which companies in the AI industry may be required to produce training and algorithm data. In Tremblay v. OpenAI, Inc., Tremblay accused OpenAI of using copywritten material to train AI products. OpenAI produced two training datasets, and Tremblay requested that the Court compel OpenAI to produce a third dataset–—the dataset used to train the GPT-4 model. Trembly argued there was no basis under the Federal Rules of Civil Procedure for the exclusion of the dataset, and specifically that the dataset was reasonably accessible without undue burden or cost under Rule 26(b)(2)(B). Joint Discovery Letter Brief filed by Paul Tremblay et. al. at 1, Tremblay v. OpenAI, Inc., 716 F. Supp. 3d 772 (N.D. Cal. 2024) (3:23-cv-03223), 1 n.1, ECF No. 254.
In response, OpenAI attempted to produce a limited dataset, arguing that the training dataset is "staggeringly enormous" and that the production would "raise a significant risk of exposing competitively sensitive data." Id. at 3. The judge did not find the burden of production to be unduly burdensome and ordered the production of the contested dataset. Aruni Soni, OpenAI to Provide Training Dataset to Authors in Suit (Correct), Bloomberg Law (Jan. 28, 2025). The court did agree with OpenAI that the production should be sealed, keeping OpenAI's training data private for the time being. Id.
Conclusions
To date, JAMS is the only arbitral tribunal to release a set of guidelines directed the resolution of disputes involving AI technology. Companies in the AI industry may wish to take advantage of JAMS's procedures and draft arbitration provisions accordingly.
Meanwhile, courts and arbitral tribunals will continue to grapple with the novel issues presented by disputes involving this technology, and legal practitioners' use of AI tools in the preparation of pleadings and other legal documents.
Originally published by New York Law Journal.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
