ACR-GNY - AI in Arbitration: Independent Research or Permissible Assistance?

Log in

Home
Learn
Resolution Roundtable Blog
AI in Arbitration: Independent Research or Permissible Assistance?
Add comment

AI in Arbitration: Independent Research or Permissible Assistance?

Monday, December 22, 2025 11:47 AM | Anonymous

How AI Could Impact International Arbitration | Arbitration Monitor

When arbitrators rely on AI to analyze the record—summaries, exhibit identification, or transcript searches—does this constitute “independent research,” and should the parties be given access to the AI outputs to ensure accuracy and fairness?

What are your thoughts?

Add comment

Comments

Tuesday, December 23, 2025 2:23 PM | Eric D. Kuritzky, Architect, CBO, Arbitrator

I don't believe in the use of AI for arbitration, especially the decisions of the arbitrator. It's a relatively simple process...each side presents their position, with back up documents or testimony. We employ our experience and knowledge to determine a conclusion from that. Aside from possibly confirming some definitions, or specific regulations, it is not our job to research. At some point, I'm sure, AAA will develop an AI to do our job...in minutes. I'll hopefully be retired by then.

Link • Reply
- Tuesday, December 23, 2025 3:59 PM | Denise Presley
  
  Eric, I couldn’t agree more. AI is independent research. I’d also add- any time saved, by using AI, is often usurped by the time spent checking its output to ensure accuracy and reliability, which I don’t believe should be billed to the parties.
  
  Link • Reply
Tuesday, December 23, 2025 2:36 PM | Geoffrey BH

AI must not be used for decision making <full stop>
It is totally dependant on the user's prompt and is valid for research if the data source can be identified.
It follows that, if a source of relevance is found, it is that, and not the AI material, which must be laid before both parties to make representation as they see fit.

Link • Reply

Author

* Comment
Tuesday, December 23, 2025 2:40 PM | Raoul East Drapeau

I use AI every day in my non-arbitration work. It is an incredible time saver when I ask it to search for relevant facts that might take me days to research on my own. On a case, to help make sure that it clearly expresses what I wanted to say, I have asked it to clean up something that I wrote. But since I have seen it make mistakes, to give it the facts on a case as I see them and then ask it to write an award - never.

Link • Reply
Tuesday, December 23, 2025 3:25 PM | G. David Miller

To the extent one should cite the authority on which you rely in your award, that authority should be cited. But there is no obligation to cite the research framework you utilize whether it is Westlaw or AI. That is your work product.

Link • Reply
Tuesday, December 23, 2025 3:45 PM | Mark Bunim

"Summaries, exhibit identification or transcript searches" does not constitute use of generative AI, but rather Document AI. Therefore, I do not believe it is "independent research," and thus see no duty to disclose.

Link • Reply
- Tuesday, December 23, 2025 6:40 PM | Paul Marrow
  
  Using AI is unauthorized research and unauthorized research is a non no, plain and simple. Document AI is not research as long as the AI is reviewing documents already produced as a part of document exchange, etc. While I personally don't use AI for anything involving arbitration, if I were to want to I would first ask the parties for permission. That's so simple and it clears away all concerns.
  
  Link • Reply
Wednesday, December 24, 2025 10:53 AM | Tom Levak

I don’t use AI to write, to do research, or for anything else. My work is my own. (It is fun to use ChatG to create poems.)

Link • Reply
Wednesday, December 24, 2025 1:56 PM | Lisa Renee Pomerantz

When I am unsure about how to decide a disputed factual (or legal) issue, my practice is to request the parties to submit briefs addressing the issues. I am reluctant to rely on AI to ensure I am considering and properly interpreting all relevant exhibits or testimony. If I were to use AI, I would give the parties an opportunity to comment on or supplement the results.

Link • Reply
Thursday, December 25, 2025 2:55 PM | Mark H. Alcott

The technology is still too new and the possibility of a mistake is still too great to use AI. And, under any scenario, anything found through that process should be put before the parties for comment before it is relied upon.

Link • Reply
Thursday, December 25, 2025 3:32 PM | Steven Skulnik

We've been using tools built into the Windows and Apple operating systems, and built into any pdf reader, to search for evidence for 25 years without even thinking there's an issue. Generative AI tools are just more powerful. As long as the arbitrator uses tools to find evidence in the record and views the evidence, without relying on the Gen AI output, there is no issue. Relying on the Gen AI output without verifying is not good practice and disclosure can't fix that.

Link • Reply
Friday, December 26, 2025 10:12 AM | Tuneen Chisolm, Esq., M.S.Ch.E.

The question posed limits the use of AI for searching to a closed universe consisting of the record, so there is no actual “research” being done. Using AI to “analyze the record—summaries, exhibit identification, or transcript searches—“ does not constitute “independent research.” As for whether the parties should “be given access to the AI outputs to ensure accuracy and fairness,” I think not. The parties are not entitled to access to the arbitrator’s digestive process, whether that be handwritten notes or AI summaries. A responsible AI user will verify results of the AI record search and summary before using them for the award. Any use of AI for the specified functions should assist but not supplant the arbitrator’s individual effort. All of that said, blind reliance on AI record search and summary seems unethical, should be disclosed, and likely would not keep an arbitrator in business.

Link • Reply
Sunday, December 28, 2025 5:58 PM | Nancy Greenwald

I have been using the Clearbrief software the AAA makes available to it's panelists as a tool for uploading and organizing pleadings, exhibits, and transcripts. I think it's a great. It decreases the time I would otherwise spend sorting through voluminous documents, and helps ensure that my document review on issues is thorough, while saving the parties money. I do not use LLM's in reviewing cases in my work as an arbitrator and mediator, but I do use ChatGPT and Claude regularly for non-case related work. I do not do independent research, using AI or any other tool.

Link • Reply