One of the most practical questions in AI-related arbitration is not whether AI can be used at all. It is when the neutral should disclose that use, and what that disclosure should actually say.
That question matters because bad disclosure can create almost as much trouble as no disclosure.
If the statement is too vague, it does not help the parties evaluate fairness, confidentiality, or due process. If it is too performative, it can turn routine workflow decisions into distraction and motion practice. The goal is not to dramatize AI use. The goal is to preserve confidence in the process.
This checklist is designed for arbitrators, institutions, and counsel thinking through neutral disclosure in AI-related proceedings.
What this checklist is for
Use it when:
- a neutral is considering using AI tools in an arbitration,
- parties want a practical framework for disclosure expectations,
- a case involves especially sensitive evidence or confidentiality concerns,
- or a procedural order needs a clean starting point.
This is practical informational guidance, not legal advice.
1. Identify who is using AI and for what purpose
Ask first:
- Is the issue the arbitrator’s own use of AI tools, the parties’ use, or both?
- Is the tool being used for clerical support, research, summarization, translation, evidence review, drafting, or something more consequential?
- Is the use occasional and low-risk, or embedded deeply enough to affect how the proceeding operates?
Not every use raises the same disclosure question. A spelling aid is not the same as a tool used to summarize evidence or generate draft reasoning.
2. Decide whether the use is material to the proceeding
A useful disclosure standard is materiality, not novelty.
Ask:
- Could the AI use reasonably matter to party confidence in fairness or due process?
- Could it affect how evidence is handled, how issues are framed, or how reasoning is developed?
- Would a reasonable party want to know about this use before the award is issued?
The AAA-ICDR’s March 2025 guidance says arbitrators should disclose generative AI use when that use materially impacts the arbitration process or the reasoning underlying their decisions. That is a better frame than blanket disclosure of every digital tool.
3. Describe the function, not just the brand name
If disclosure is warranted, do not stop at naming a product.
The better disclosure explains:
- what the tool was used for,
- what it was not used for,
- whether outputs were independently reviewed,
- and whether any confidential case information was entered into the system.
That gives the parties something useful to evaluate.
4. Address confidentiality directly
Confidentiality is one of the fastest ways AI use can become a procedural problem.
Ask:
- Did the tool receive party names, case specifics, evidence, prompts, draft rulings, or other protected information?
- Were those inputs handled in a secure environment?
- Do the tool’s data practices create avoidable risk?
California Standard 15 and the AAA-ICDR guidance both point in the same direction: confidential arbitration information should not be handled casually, and arbitrators should not use tools that fail to protect sensitive case material.
5. Confirm that human judgment stayed with the neutral
Disclosure should make clear that the tool assisted the work but did not replace judgment.
Ask:
- Did the tool merely support workflow, or did it meaningfully shape analysis?
- Who verified accuracy, citations, and factual characterizations?
- Who retained final control over findings, reasoning, and conclusions?
Ciarb’s 2025 guideline and the AAA-ICDR’s 2025 guidance both reinforce the same principle: AI may support the process, but the neutral remains responsible for the decision.
6. Check for hidden conflicts and platform relationships
AI disclosure is not only about prompts and summaries.
It can also raise questions about:
- vendor relationships,
- institutional tool mandates,
- consulting ties,
- proprietary tool development,
- or any financial or professional interest that may intersect with the case.
California Standard 7 is broader than AI, but that is exactly why it remains useful. A neutral should ask whether any current relationship, prospective relationship, or professional arrangement connected to the tool environment could reasonably matter to perceptions of impartiality.
7. Consider whether a standing protocol is better than a one-off disclosure
Some cases are too sensitive for ad hoc handling.
In those matters, it may be better to establish a protocol that addresses:
- whether AI tools may be used at all,
- which categories of use require disclosure,
- whether party consent is required for certain uses,
- confidentiality restrictions,
- and whether AI-assisted work product must be verified in a specified way.
This can be cleaner than improvising after concerns surface mid-case.
8. Make the disclosure early enough to matter
A disclosure that arrives only after an award is drafted may be too late to do useful work.
Ask:
- Can the issue be surfaced during appointment, early scheduling, or the first procedural conference?
- If the use evolves later, is there a continuing duty to update the parties?
California Standard 7 expressly recognizes a continuing duty of disclosure. That concept translates well to AI use if the neutral’s practice changes in a way that becomes materially relevant.
9. Keep the record clear and proportionate
Once a disclosure is made, memorialize it cleanly.
That can include:
- a written disclosure statement,
- a procedural order,
- a conference transcript entry,
- or agreed guidance on tool use and data handling.
The point is not to create ceremonial paperwork. It is to avoid later fights about who knew what and when.
10. Watch for red flags
Disclosure likely needs more attention if:
- the neutral used generative AI on confidential submissions,
- the tool influenced reasoning or credibility assessment,
- the neutral cannot explain how outputs were verified,
- the parties were not told until late in the case,
- or the tool environment creates unresolved data protection concerns.
Those are not merely optics problems. They can become enforceability, fairness, and legitimacy problems.
Sample practical disclosure elements
In many cases, a useful disclosure will answer five questions:
- What tool category was used?
- For what limited purpose was it used?
- What confidential information, if any, was entered?
- What verification steps were taken?
- What remained exclusively within the neutral’s own judgment?
That structure is usually more valuable than a generic statement that “technology was used.”
FAQ
Does every arbitrator use of AI require disclosure?
No. The better question is whether the use is materially relevant to fairness, confidentiality, due process, or the reasoning process.
Is disclosure only about generative AI?
No. Generative AI creates obvious issues, but disclosure questions can also arise from translation, analytics, summarization, or evidence-handling tools.
Should disclosure include the exact product name?
Sometimes, but the function usually matters more than the brand. Parties need to understand what the tool did and what risks it may have created.
Can parties address this by agreement?
Yes. Many cases will benefit from a procedural order or early agreement on AI use, disclosure, confidentiality, and verification.
What is the biggest mistake?
Treating disclosure as a publicity exercise instead of a fairness and process-design issue.
Conclusion
An AI neutral disclosure checklist is valuable because it keeps the process credible before suspicion hardens into conflict.
The real objective is not to turn every technical detail into a dispute. It is to make sure that when AI touches an arbitration, the parties know enough to evaluate the process, protect confidentiality, and trust that human judgment stayed where it belongs.
Further Reading
- California Standard 7. Disclosure: https://courts.ca.gov/cms/rules/index/ethics/ethics7
- California Standard 15. Confidentiality: https://courts.ca.gov/cms/rules/index/ethics/ethics15
- AAA-ICDR Guidance on Arbitrators’ Use of AI Tools, March 2025: https://www.adr.org/media/g1fgccns/2025_aaa-icdr-guidance-on-arbitrators-use-of-ai-tools-2.pdf
- Ciarb Guideline on the Use of AI in Arbitration, updated September 2025: https://www.ciarb.org/media/bpndtcgu/guideline-on-the-use-of-ai-in-arbitration_updated-sept-2025.pdf
