Labor & Employment Law Daily Wrap Up, NAA CORNER—NAA issues new ethical guidelines on arbitrators’ use of artificial intelligence tools, (Apr 21, 2026)
Organizations Mentioned:National Academy of Arbitrators
By Andrea Dooley, NAA Member
These guidelines provide much-needed direction for arbitrators on how existing ethical obligations apply to growing technology and for practitioners who are concerned about how arbitrators intend to use AI tools in their practice.
Late last month, the National Academy of Arbitrators’ Committee on Professional Responsibility and Grievance issued new guidance on how and when labor arbitrators can ethically use generative artificial intelligence or “Generative AI” in their practice. “Opinion No. 28: The Arbitrator’s Use of AI” explains which sections of the NAA Code of Professional Responsibility (“CPR”) apply to Generative AI and what labor arbitrators should consider and disclose if they elect to use Generative AI. This signals that the use of AI tools will be subject to existing ethical obligations particularly when it comes to disclosure, confidentiality and decision-making authority.
While the NAA Code of Professional Responsibility applies to members of the National Academy of Arbitrators, the Federal Mediation and Conciliation Service and the American Arbitration Association also require adherence to the CPR for members of their labor arbitration panels.
Familiarity with hazards. The Opinion identifies four areas of the CPR that likely apply to the use of Generative AI and limits its use by arbitrators. Section 1 of the Code concerning arbitrator’s qualifications and responsibilities to the profession requires that arbitrators be familiar with the hazards associated with Generative AI, including the likelihood of “hallucinations” or false information as well as the possibility of bias or prejudice in the tool’s output. This places the responsibility on arbitrators to understand the capabilities, but also be mindful of their limitations before using them.
Disclosure and consent. Section 2 of the CPR concerns the arbitrator’s obligations concerning disclosure and consent. Opinion No. 28 asserts that Section 2 requires that arbitrators disclose their use of Generative AI to the parties at the earliest opportunity because “a party could perceive the use of [a] GAI tool as a source of outside influence on an arbitrator, or a delegation of the decision-making process to a GAI tool.” Arbitrators do not need to disclose the use of word processing software for formatting, proofreading, calculating, or other tasks in programs which do not use or retain the arbitrator’s input for other purposes.
Privacy. Section 2 of the CPR also requires “Privacy of Arbitration,” which prohibits an arbitrator from making public an award without the consent of the parties. To the extent that any part of the record, including the transcript, exhibits or award, might be used to train an AI tool, the arbitrator must have the consent of the parties for that use. “An arbitrator should not submit ‘any significant aspect’ of an arbitration case record or related data or any identifying information about the parties for analysis by an AI tool unless the arbitrator is confident that the submission will not subject that material to potential disclosure to non-parties.”
A separate concern exists about the possibility that the owner or operator of the AI tool could be subject to a third-party subpoena. While Opinion No. 28 does not dive into the particulars of this issue, it does warn arbitrators about the retention of information by an AI tool, including an arbitrator’s requested outputs, which might be later viewed by a reviewing court as a delegation of the decision-making authority of the arbitrator.
Further, federal courts do not recognize a per se attorney-client privilege for materials generated by AI. See U.S. v Heppner, 25-CR-503 (Rakoff, Southern District of N.Y., U.S. District Court). Courts may be reluctant to enforce awards where the arbitrator relied on AI-generated material in reaching their decision and are unlikely to consider AI-generated material to be covered by any privilege an arbitrator may assert.
Use of assistants. Section 2 of the CPR also covers the use of assistants, prohibiting an arbitrator from delegating any decision-making function to another person without the consent of the parties. Opinion No. 28 extends that provision to non-person assistants such as Generative AI tools that arbitrators might use, including the use of these tools to draft fact statements. Advisory Opinion No. 12 states: “It is essential to remember that decision making starts with fact finding. The parties rely on the arbitrator to determine what the facts are (credibility) and which ones are important (weight). [They] should not delegate those functions without the parties’ knowledge.” AI tools raise similar concerns regarding an arbitrator’s independence and authorship.
Related expenses. In addition to requiring disclosure about the use of Generative AI tools in the decision-making process, arbitrators must disclose to the parties if they intend to bill the parties for the expenses related to Generative AI tools “prior to appointment.” This is extension of the transparency that arbitrators are expected to have toward the parties appearing before them.
Permissible uses. Finally, the Committee on Professional Responsibility and Grievances also identifies permissible uses of the tool if its use is consistent with the CPR. An arbitrator can use a Generative AI tool to assist with legal research as long as the arbitrator carefully reviews the output for accuracy and completeness. An arbitrator can use an AI tool to aid in error-correction of an award as long as it does not involve substantive changes to the award. Similarly, an arbitrator could use an AI tool to find omissions or inconsistencies in the award as long as the arbitrator verifies the output before relying on the AI tool to make changes to their award.
These guidelines provide much-needed direction for arbitrators on how existing ethical obligations apply to growing technology and for practitioners who are concerned about how arbitrators intend to use AI tools in their practice. These guidelines will evolve along with the AI tools that are available but will continue to depend on the Code of Professional Responsibility to ensure that labor arbitrators are adhering to the highest standards of professional practice.
About the author. Andrea Dooley has served as an arbitrator for grievances and arbitrations, fact-finding hearings, civil service hearings, mediation, labor-management partnerships, and interest-based bargaining and has authored numerous newsletters and publications.
Companies: National Academy of Arbitrators
MainStory: Arbitration Labor IndustryNewsTrends AINews Privacy