On 23 April 2024, JAMS announced its Artificial Intelligence Disputes Clause and Rules (“JAMS AI Rules”), effective 15 April 2024.  According to JAMS, these Rules “reflect the latest developments and trends in the ADR space and address the rise in usage and development of AI systems and smart contracts.”

Today, artificial intelligence (“AI”) is a worldwide hot topic, including within arbitration.  As the use of AI inevitably grows over the coming years, disputes concerning its use will undoubtedly arise.  Currently, novel cases are being brought before the U.S. courts relating to big technology companies’ use of generative AI.  For example, visual artists recently filed a class action against Google, alleging that it improperly used copyrighted images to train “Imagen,” its AI system that converts text to images.

JAMS, cognizant of the impending rise in AI-related arbitrations, is the first arbitral institution to publish a set of arbitration rules that are “tailored to the complexities of AI.”  JAMS describes these new rules as “clear guidelines and procedures that address the unique challenges presented by AI, such as questions of liability, algorithmic transparency, and ethical considerations.”

The JAMS AI Rules are new arbitration rules intended to govern disputes involving AI.  They do not provide rules or guidance on the use of AI in arbitrations.  Rather, they are a set of rules designed to “refine and clarify procedures for cases involving AI systems.”

The JAMS AI Rules are, in many respects, identical to the JAMS Comprehensive Arbitration Rules (“JAMS Comprehensive Rules”), though there are important differences in the JAMS AI Rules that are intended to offer parties an alternative to the JAMS Comprehensive Rules for disputes concerning AI systems.  This post summarizes and analyzes some of the key provisions in the JAMS AI Rules that differ from the JAMS Comprehensive Rules, including: (i) the definition of AI; (ii) procedures relating to the production and inspection of AI systems and related materials, including expert evidence on those systems; (iii) mandatory expedited procedures; and (iv) additional safeguards to protect confidential information.

 

Definition of Artificial Intelligence

Rule 1(e) of the JAMS AI Rules broadly defines “Artificial Intelligence” or “AI” as “a machine-based system capable of completing tasks that would otherwise require cognition.”  Many machine-based systems are potentially captured by this definition, beyond those that are today most typically considered to constitute AI.  In essence, whether a machine-based system constitutes AI under the JAMS AI Rules depends on what tasks “otherwise require cognition.”  While that definition undoubtedly would cover sophisticated machine-based systems such as ChatGPT, it might also cover rudimentary machines that the drafters of the Rules may not have intended.  For example, arguably computers perform many tasks that would “otherwise require cognition,” such as the short term or long-term storage of information.

The breadth of this definition impacts parties that incorporate the JAMS AI Rules into their agreements.  For example, as discussed below, the JAMS AI Rules impose significant restrictions on the production and inspection of “any AI systems or related materials.”  The breath of the Rules’ definition of AI may result in an unintentionally broad application of these restrictions.  For example, parties might argue that machine-based systems within their organization, such as internal computer systems, constitute AI under the JAMS AI Rules and thus should be afforded the heightened protections afforded to “AI systems or related materials.”

Similarly, as discussed below, the JAMS AI Rules require parties to mutually agree upon experts providing opinions on “AI systems, including any expert(s) inspecting any AI systems or related materials,” absent which the tribunal will designate such experts.  The broad definition of AI therefore may result in ambiguity as to which issues may be addressed by party-appointed experts.

In contrast, in its recently published Guidelines on the Use of Artificial Intelligence in International Arbitration, the Silicon Valley Arbitration and Mediation Centre (“SVAMC”) defines AI as “computer systems that perform tasks commonly associated with human cognition, such as understanding natural language, recognising complex semantic patterns, and generating human-like outputs.”  This definition similarly leaves room for dispute as to what tasks are “commonly associated with human cognition.”  However, it also helpfully provides examples of such tasks that are akin to what is considered today to constitute AI, which may give tribunals more scope to determine that more rudimentary machines do not fall within its scope.  Ultimately, the breadth of both the JAMS and SVAMC definitions of AI reflects AI’s complexity, further highlighting the need for carefully constructed procedures to govern disputes arising pursuant to AI.

 

Production and Inspection of Artificial Intelligence Systems

The JAMS AI Rules provide a novel process for the production and inspection of “any AI systems or related materials” as a safeguard to protect the sensitive and confidential information and trade secrets inherently intertwined with AI systems.  Rule 16.1(b) provides:

“The production and inspection of any AI systems or related materials, including, but not limited to, hardware, software, models and training data, shall be limited to the Disclosing Party making such systems and materials available to one or more expert(s) in a secured environment established by the Disclosing Party. The expert(s) shall not transmit or remove any produced materials or information from such environment.”

Practically, this means that a party requesting the disclosure of the opposing party’s “AI systems or related materials” will not gain access to that information, nor will its counsel. Instead, the party will have to rely on the expert(s) that review such information in the secured environment established by the disclosing party.

The scope of the types of information that may be subject to this restriction is vague, and possibly overbroad.  For example, it is unclear what “related materials” to AI systems constitutes.  Taking ChatGPT as an example, the underlying software running the system would appear to fall within the scope of this protection.  However, it is unclear whether internal communications at OpenAI (the owner of ChatGPT) regarding ChatGPT would constitute “related material” and thus also fall within the scope of these protections.  This ambiguity might lead to procedural arguments between parties as to which documents and information can reasonably be withheld from disclosure to the opposing side and its counsel.

Additionally, Rule 16.1(b) requires that:

Any expert(s) providing opinions on AI systems, including any expert(s) inspecting any AI systems or related materials, shall be mutually agreed upon by the Parties.  If the Parties are unable to agree, the Arbitrator shall designate such expert(s).  The Arbitrator shall first attempt to designate such expert(s) from a list of third-party experts maintained by JAMS, subject to availability and appropriate qualifications.”

On its plain terms, this provision would appear to prevent parties from unilaterally appointing experts to provide opinions on AI systems, or to inspect AI systems or related materials under the disclosure process discussed above.  It is unclear whether this leaves available the option for a party to unilaterally appoint an expert to provide an opinion on AI-related issues.  The provision does not prohibit a party from appointing an expert to opine on issues of AI more generally, but it appears to prohibit unilateral appointment of an expert to opine on or inspect the specific AI systems (or related materials) at issue in the arbitration.

Party-appointed experts are often a crucial element in arbitration, especially with respect to technical issues.  It is often important for parties and counsel to discuss technical issues with their appointed experts in the preparation of expert reports, parties’ briefs, and for the hearing.  It is unclear how, with the restrictions on expert opinions imposed by Rule 16.1(b), this process will unfold under the JAMS AI Rules.

Further, it may be difficult for parties to reach mutual agreement on the appointment of experts to provide opinions on, or inspect, AI systems.  Consequently, tribunals may often be called upon to designate such experts.  The qualifications and expertise of the experts, and the number of such experts on the “list of third-party experts maintained by JAMS,” also remain to be seen.

Finally, Rule 16.1(b) provides that:

“[e]xpert testimony on any technical issues related to an AI system shall be limited to a written report requested by the Arbitrator addressing questions posed by the Arbitrator, and testimony at the Hearing of such expert(s).”

This limitation appears to further curtail the parties from putting forward expert evidence that they deem important, instead requiring the tribunal to have sufficient knowledge of the technical issues to ensure that the right questions are being put to the expert(s).  Additionally, it is not clear what would constitute a “non-technical” issue related to an AI system on which a party arguably could present expert testimony.  Nevertheless, as discussed above, a party would still appear to be restricted from appointing unilaterally an expert to provide a “non-technical” opinion on “AI systems,” as the provision appears to require that such experts must either be mutually agreed or designated by the tribunal.

 

Mandatory Expedited Procedure

As discussed above, the JAMS AI Rules and the JAMS Comprehensive Rules are largely identical, subject to several differences in the JAMS AI Rules designed to be tailored to AI disputes.  One of these differences is that per Rule 16.1(a) of the JAMS Comprehensive Rules, the parties are provided with the option to apply a set of “expedited procedures,” whereas the JAMS AI Rules automatically applies those expedited procedures as standard procedures.  Rule 16.1 of the JAMS AI Rules (“Procedures”) is almost identical to Rule 16.2 of the JAMS Comprehensive Rules (“Where Expedited Procedures Are Applicable”).

These procedures in the JAMS AI Rules require, inter alia:

  • The production and inspection of AI systems by a mutually agreed upon expert (or tribunal-appointed expert), as discussed above.  This, as well as the inclusion of the AI Disputes Protective Order (discussed below), is the only difference between the procedures in the JAMS AI Rules and the expedited procedures in the JAMS Comprehensive Rules.
  • The limitation of one discovery deposition per side, unless determined, “based on all relevant circumstances, that more depositions are warranted.”
  • The resolution of discovery disputes on an “expedited basis,” meaning that, inter alia: (i) discovery decisions will likely be made by the chair (or another member of the arbitral tribunal) alone; and (ii) “lengthy briefs on discovery matters should be avoided,” with “brief letters” being sufficient in most cases to inform the arbitrator of the issues to be decided.
  • “[P]ercipient” discovery i.e., discovery pertaining to issues of fact, must be completed within 75 days, and expert discovery within 105 days, of the preliminary conference.
  • The hearing shall commence within 60 days after the cutoff for percipient discovery, resulting in only 135 days between the preliminary conference and the hearing.

These expedited procedures are a significant step towards streamlining arbitrations, making them more cost and time efficient.  However, such streamlining may curtail parties’ ability to fully present their cases, especially where novel and complex issues involving AI may be at stake.  It is worth noting that the JAMS AI Rules provide the tribunal with the power to alter any of the expedited procedures “for good cause,” which may negate the potential risks that could arise by wholesale adoption of the expedited procedures.

 

Increased Confidentiality

As discussed above, the JAMS AI Rules provide safeguards for the production and inspection of “AI systems and related materials” that restrict disclosure of this information.

The JAMS AI Rules provide further safeguards to protect confidentiality.  For example, Rule 16.1(a) provides for the application of a protective order at Appendix A, which “shall apply to protect confidential information, unless another form of protective order is agreed to by the Parties.”  The protective order, similar to protective orders often voluntarily entered into by parties in arbitrations, allows each party to designate documents or other material as either “Confidential” or “Highly Confidential,” restricting the other party’s use and dissemination of such materials.

Furthermore, Rule 26(b) provides that:

“All Parties to the Arbitration and their counsel shall strictly maintain in confidence all details of the Arbitration and the Award, including the Hearing, except as necessary to participate in the Arbitration proceeding and the Hearing, in connection with judicial challenge to or enforcement of a decision, or unless otherwise required by law or judicial decision.”

 

Conclusion

The JAMS AI Rules are a step forward from arbitral institutions to prepare for the inevitable rise of disputes arising out of the use of AI.  Enhanced confidentiality safeguards will be welcomed by parties that are concerned about the disclosure of highly sensitive, confidential information, such as details of their internal AI systems and the data those systems use and generate.  Further, attempts to streamline the arbitral process – including short timelines for production of evidence and for the hearing, as well as limiting AI-related expert evidence – may also be welcomed by parties seeking quicker resolutions of disputes at lower costs.

However, the broad definition of AI in the JAMS AI Rules may result in an unintentionally broad application of the Rules to machine-based systems that the drafters did not intend to fall under the heightened protections under the Rules with respect to documents and material.  Such a broad application might hamper the efficient resolution of some disputes, by limiting parties and their counsel from reviewing documents and material, and providing expert opinions on such information, as they would normally be permitted to do under the JAMS Comprehensive Rules (and other arbitral rules).

Furthermore, the future of AI, including the disputes that will inevitably arise from its growing use, is still unknown.  As the commercial usage of AI increases, it is likely that novel, complex and technical disputes will follow.  Certain provisions of the JAMS AI Rules, such as the inability of a party to present its own independent expert to opine on AI systems, or inspect AI systems and related materials, may hinder parties’ ability to fully put forward their cases, and prevent tribunals and counsel from fully understanding the complexities of the issues at stake.  With that in mind, parties should carefully consider whether to incorporate the JAMS AI Rules into their agreements, while ensuring that they have the important confidentiality safeguards in place in disputes arising out of AI systems.


________________________

To make sure you do not miss out on regular updates from the Kluwer Arbitration Blog, please subscribe here. To submit a proposal for a blog post, please consult our Editorial Guidelines.


Kluwer Arbitration
This page as PDF

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.