The more practitioners use AI, the more they appear to trust it.
That’s one of the findings from The State of AI in U.S. Arbitration 2026, a new study by the American Arbitration Association (AAA) and Jus Mundi. The research suggests that, as practitioners gain experience with AI, they are developing more nuanced views of where it can add value.
Among those surveyed, daily AI users reported an average trust score of 3.42 out of 5, compared with 0.83 among respondents who have never used AI and do not plan to.
But greater familiarity does not translate into the same level of trust for every task.
Trust Depends on the Task
Trust in AI is not one-size-fits-all.
“It’s trust for what? If it’s trust for reading the cases, that’s a very low level of trust. But if I’m asking AI to alphabetize something, I don’t need to trust it, because the output is so easy to check,” said Jack Levin, arbitrator and mediator at Levin ADR and a participant in the report’s qualitative interviews.
For now, respondents appear most comfortable using AI to review, organize, and analyze information. Fifty-one percent of respondents said they use or would use AI for document review and organization, while 46% said the same for legal research and precedent analysis.
What Wider Adoption May Require
When asked what would accelerate the adoption of responsible AI in arbitration, respondents prioritized clearer guidance and more specialized tools. Institutional guidelines ranked first at 30%, followed by domain-specific tools at 24%.
This aligns with the concerns respondents identified elsewhere in the study. Inaccuracy or hallucinations were the most frequently cited concern at 78%, followed by confidentiality and data security at 48%.
When it comes to inaccuracy, experienced users appear to manage this risk through verification rather than avoiding AI altogether. For John Hanish, partner at Maynard Nexsen PC and a participant in the report’s qualitative interviews, using AI for legal research requires practitioners to verify the output before relying on it:
“They go back and check, make sure the cases are really cases, that everything the tool gave you actually exists and isn’t just made up,” he said.
Explore the Full Report
These findings represent just part of the picture of how AI is being used, understood, and governed across U.S. arbitration today. Download The State of AI in U.S. Arbitration 2026 for the full findings and analysis.