AI in Arbitration: When the Tool Turns a Molehill Into a Mountain

Artificial intelligence is already changing arbitration. Some of that change is useful. AI can help lawyers organize records, identify duplicate documents, build timelines, check citations, summarize transcripts, and prepare more efficiently. Used carefully, it can reduce cost and improve focus.

But there is another side to the story.

AI can also make arbitration harder. Not because artificial intelligence is inherently bad, and not because thoughtful AI tools have no place in dispute resolution. The concern is more practical: parties and counsel can use AI in ways that generate more claims, more arguments, more documents, more factual disputes, and more process fights than the case actually warrants.

Put differently, AI has a remarkable ability to turn a molehill into a mountain.

The answer is not to ban AI or pretend it is a passing fad. It is to manage it early, set expectations, protect confidentiality, verify what matters, and keep the process proportional to the dispute.

In other words, AI should help arbitration become more focused. Too often, unmanaged AI does the opposite.

The Promise Is Real, But So Is the Problem

A recent AAA article captured the broader trend well: “Technology is making disputes more common, more complex, and harder to control.” It noted that technology now sits at the center of operations across industries and, increasingly, at the center of disputes. It also observed that AI makes demand letters and pleadings easier to generate, leading to “more disputes, arriving faster, and often with less friction at the outset.” 

That point resonates. Arbitration is supposed to be flexible, efficient, and tailored to the dispute. But AI can change the economics of dispute creation. A demand letter that once required substantial lawyer time can now be drafted quickly. A party with a modest complaint can generate a sophisticated-looking legal theory. A lawyer can produce more arguments, more summaries, and more proposed findings because the tool makes volume easier.

Volume, however, is not the same as value.

In arbitration, the goal is not to say everything that can be said. The goal is to help the arbitrator understand what matters. AI can assist with that. But without human discipline, it can make a small issue look bigger, more complex, and more legally significant than it really is.

That is how the molehill becomes the mountain.

Managed AI Is Different From Unmanaged AI

It is important to be fair about this. There are responsible, structured uses of AI in ADR.

The AAA’s AI Arbitrator program, for example, is described as an opt-in process for certain documents-only cases. The AAA emphasizes that the process involves “final decisions issued by human arbitrators” and that the AI tool does not replace the arbitrator’s judgment. Similarly, the AAA-ICDR’s work with ClearBrief reflects a structured use of technology to assist with organization, citation verification, and efficiency.

Those examples are not the problem. They involve defined tools, known procedures, confidentiality considerations, disclosure, and human review.

The bigger concern is unmanaged AI use by parties and counsel: uploading sensitive materials into unknown systems, generating submissions that are not checked, turning every possible argument into a brief, and creating new disputes about what was generated, what was verified, what is confidential, and what is real.

The issue is not AI itself.

The issue is AI without judgment.

AI Makes It Easier to Overstate a Case

Good advocacy requires selection. Lawyers decide which facts matter, which arguments are worth making, and which points should be left alone. AI can weaken that discipline when it is treated as a substitute for judgment.

A party may ask AI to identify every inconsistency in a deposition. It will find some. It may also identify trivial differences that mean nothing. A lawyer may ask AI to draft claims from a set of facts. It may generate a plausible complaint. It may also dress up a routine contract disagreement as fraud, unfair competition, breach of fiduciary duty, and conspiracy.

That creates real process costs.

The opposing party must respond. The arbitrator must sort through it. Discovery expands. The case becomes more expensive. Settlement becomes harder because inflated theories create inflated expectations.

The danger is not that AI produces nothing useful. The danger is that it produces too much that sounds useful.

In the hands of an undisciplined advocate, AI does not narrow the case. It enlarges it.

Verification Becomes Part of the Cost

The hallucination problem is not theoretical. Stanford researchers have found that public-facing large language models hallucinate at high rates when asked legal questions. A separate Stanford evaluation of specialized legal research tools found that even legal AI products still produced incorrect information at meaningful rates.

That matters in arbitration because arbitration is submission-driven. Arbitrators rely on briefs, witness statements, chronologies, expert reports, exhibit summaries, damages calculations, and proposed findings. If AI-assisted work product contains inaccurate citations, misstated holdings, distorted testimony, or overconfident factual summaries, someone must catch it.

That “someone” is usually opposing counsel, the arbitrator, or both.

So AI may save time on the front end while creating verification costs on the back end. The lawyer who submits AI-assisted work may have spent less time drafting it. But if the submission is unreliable, everyone else spends more time checking it.

That is not efficiency. It is cost-shifting.

And again, the case grows. What should have been a narrow dispute over the facts becomes a broader dispute over the accuracy of summaries, the reliability of citations, and the trustworthiness of the submission itself.

The molehill gets bigger.

Digital Evidence Is Becoming Easier to Challenge

AI is also changing the way parties think about evidence.

Courts and arbitration users are beginning to confront a basic problem: digital evidence is becoming easier to create, alter, enhance, or challenge. Screenshots, text messages, emails, audio clips, video clips, social media posts, spreadsheets, customer records, and platform data may all become subjects of authenticity disputes.

The exhibit may be real. But the cost of proving it is real may increase.

That can turn ordinary evidentiary issues into side proceedings. Was the screenshot altered? Is the audio authentic? Is the image enhanced? Is the spreadsheet complete? Were the records exported from the system or recreated? Is there metadata? Is there a native file? Is expert review needed?

AI may not create the underlying dispute. But it can make the dispute harder to manage.

A single document, screenshot, or recording can become its own procedural mountain.

Confidentiality Needs More Attention

Confidentiality has long been one of arbitration’s perceived advantages. AI complicates that advantage.

The 2025 Queen Mary University of London and White & Case International Arbitration Survey found that 90% of respondents expect to use AI for research, data analytics, and document review. The principal drivers were saving time, reducing cost, and reducing human error. But the leading obstacles were errors and bias at 51%, confidentiality risks at 47%, lack of experience at 44%, and regulatory gaps at 38%. 

That tension is exactly what many arbitration users now face. AI tools can help review and organize information. But arbitration materials often include trade secrets, financial data, employment records, medical information, proprietary pricing, non-public business plans, settlement communications, or personal information.

Once sensitive material is uploaded into the wrong tool, the problem may already exist.

That does not mean parties can never use AI with confidential materials. It means counsel should understand the tool, the data settings, the retention policy, the security protections, and the confidentiality implications before using it.

This is another way AI can enlarge a dispute. A case that began as a business disagreement can become a fight over whether confidential material was improperly disclosed into an AI system. A discovery issue can become a data-security issue. A document-review shortcut can become a procedural motion.

The tool that was supposed to save time becomes the subject of the next dispute.

The Process Should Be Addressed Early

The solution is practical case management.

At the preliminary conference, the arbitrator and counsel should consider whether AI needs to be addressed. Not every case requires a formal AI protocol. A small, simple case should not become burdened by technology procedures that cost more than they save. But in cases involving sensitive information, large document sets, digital evidence, expert analysis, or allegations that AI was used to create or alter evidence, the topic should not be ignored.

A reasonable discussion can cover several points.

Will parties use generative AI with confidential materials? Are there limits on what may be uploaded? Must legal citations and quotations be verified against original sources? Must factual summaries be checked against the record? Will AI-generated or AI-altered evidence be disclosed? Will native files and metadata be preserved? Are prompts and outputs discoverable, or only in defined circumstances? Should expert reports disclose material AI use?

Those questions are not meant to create a new battleground.

They are meant to prevent one.

Use AI to Reduce the Mountain

The better use of AI in arbitration is not to generate more. It is to focus better.

AI can help organize exhibits. It can identify duplicates. It can build a draft chronology. It can flag missing documents. It can help counsel prepare a concise issue list. It can help a client understand the factual record. It can help lawyers write more clearly.

But the human lawyer must decide what matters. The human arbitrator must manage the process. The human client must understand the risks.

The best AI-assisted advocacy will not be the longest brief, the most elaborate timeline, or the thickest exhibit summary. It will be the submission that uses technology to sharpen judgment rather than bury it.

Counsel should ask a simple question before using AI output in arbitration: does this make the case clearer, or just larger?

If the answer is “larger,” leave it out.

Conclusion

AI has a place in arbitration. Used well, it can reduce cost, improve organization, and make complex records easier to understand. Used poorly, it can generate claims faster than they can be evaluated, inflate modest disputes, create verification burdens, complicate confidentiality, and turn digital evidence into a fight about authenticity.

That is not an argument against AI. It is an argument for judgment.

Arbitration works best when the process is proportionate to the dispute. AI should be held to the same standard. Use it where it helps. Control it where it creates risk. Verify it where accuracy matters. Limit it where it adds volume without value.

The future of arbitration will not be AI-free. Nor should it be.

But if AI is going to improve arbitration, parties and counsel must use it to clarify the dispute, not enlarge it. Otherwise, the tool that promised efficiency may become one more reason a case becomes harder, slower, and more expensive than it needed to be.

AI should help reduce the mountain to a molehill.

It should not turn the molehill into the mountain.

Nationwide ADR® provides arbitration, mediation, and other dispute-resolution services designed to bring structure, judgment, and clarity to demanding disputes.

Unlocking Solutions for Demanding Cases®.