The Case For Keeping AI Disputes Out Of Court

Arbitration can protect training data, models, and algorithms from the forced disclosure that litigation makes more likely.

business: the training datasets, model architectures, and algorithms that define how the system behaves. Courts have shown limited sympathy for defendants’ objections that this is too sensitive or too burdensome. In a recent case, OpenAI was ordered to turn over a contested training dataset despite arguing that the data was competitively sensitive and difficult to extract — though the court did require the data to be kept under seal.

Arbitration is usually faster and more private than court. But JAMS, one of the largest dispute resolution bodies in the country, has developed rules designed specifically for AI disputes, and they address the exact exposures that make litigation so risky.

The rules create built-in confidentiality tiers for sensitive technical materials, so protection doesn’t have to be fought for document-by-document once a dispute is underway. When AI systems need to be examined, that review happens in a secure environment, by qualified experts, under conditions the parties can rely on from the start. And because JAMS’ AI rules incorporate expedited procedures as a default, disputes get resolved in months rather than the nearly three years it takes the average federal civil case to reach resolution.

But there’s a catch: Arbitration must be written into contracts before anything goes wrong. The clause in a company’s commercial agreements, licensing deals, and employment contracts determines which forum it will be in — and most companies continue to agree to boilerplate that was drafted before anyone had thought carefully about what an AI company stands to lose in court.

Reviewing that clause isn’t a legal housekeeping task. For companies whose competitive position depends on keeping their models and data proprietary, it’s a risk management decision — and one that’s easier to make now than after the fact.

* * *

This piece draws on “Arbitrating to Protect AI: Considerations for AI Companies When Adopting Arbitration Clauses,” by Alexandra Valenti, James Breen, Timothy Keegan, and Jessica Pauley, published in the New York Law Journal. The full article includes a detailed comparison of JAMS, American Arbitration Association, and Silicon Valley Arbitration & Mediation Center guidelines as well as an analysis of how courts have approached AI-related discovery.