AI dispute resolution mixes technical language, legal language, and procedural language. This glossary defines the most important terms in plain English so the rest of the Sherafy section is easier to navigate.
AI dispute resolution
The broader category of processes used to resolve disputes involving AI systems, AI-related evidence, or AI use during the dispute process itself. It includes arbitration, mediation, litigation, expert determination, and hybrid procedures.
AI arbitration
Arbitration used to resolve disputes involving AI-related products, systems, contracts, evidence, or procedural questions.
Arbitration clause
A contract provision that tells the parties whether future disputes will be resolved through arbitration and, sometimes, how that arbitration will be conducted.
AI arbitration clause
An arbitration clause drafted with AI-specific dispute risks in mind, such as technical expertise, confidentiality, evidence handling, audit rights, or tool-use concerns.
Arbitrator
A neutral decision-maker in an arbitration proceeding.
AI tool
Any software system that uses artificial intelligence techniques to generate, classify, predict, summarize, analyze, or otherwise assist with tasks.
Generative AI
AI tools that generate content such as text, images, code, summaries, or analysis in response to prompts or other inputs.
Prompt
The instruction or input given to an AI tool. In disputes, prompts can matter because they may shape outputs, reveal intent, or expose sensitive information.
Output
The response or result produced by an AI tool.
Prompt chain
A sequence of prompts and follow-up interactions used to reach a result. In some disputes, the chain matters more than a single prompt viewed in isolation.
Model
The underlying AI system that generates or supports outputs. In disputes, questions may arise about which model version was used, how it was configured, or what limitations it had.
Model version
The specific iteration or release of an AI model at a given time. Version history may matter because behavior can change across updates.
Model licensing dispute
A dispute involving rights to access, use, fine-tune, resell, restrict, or rely on an AI model or model-based service.
Training data
The data used to train or refine an AI model. Training-data disputes may involve ownership, permission, provenance, privacy, scope of use, or competitive misuse.
Fine-tuning
A process for adapting a model using additional data or instructions to improve performance on specific tasks or within specific domains.
Retrieval
The process of pulling in external information or documents to help an AI system generate a response. Retrieval-related records can become important when a dispute turns on what sources influenced an output.
Hallucination
An output that is false, unsupported, or fabricated but presented with confidence. In disputes, hallucination is often less important as a buzzword than as a reliability problem.
AI evidence
The records and technical materials used to prove what an AI system did, how it was used, what changed over time, and what consequences followed.
Audit trail
A record showing actions, access, changes, or system events over time. In AI disputes, audit trails may help establish timing, responsibility, and system state.
Log
A structured record of system activity, events, usage, or access.
Incident report
A record documenting a failure, complaint, safety issue, operational problem, or escalation connected to an AI system or workflow.
Chain of custody
The documented path showing how evidence was collected, preserved, transferred, and handled. This can matter when parties challenge authenticity or completeness.
Authenticity
The question of whether a piece of evidence is what the party says it is.
Reliability
The question of whether a source, record, or output can be trusted for the use being made of it.
Procedural fairness
The principle that parties should receive a fair process, including a meaningful opportunity to present their case and respond to the other side.
Due process
A broader fairness concept often invoked when the integrity of the procedure itself is at issue.
Equality of arms
The idea that parties should have a fair opportunity to present their case without one side gaining an improper procedural advantage.
Independent judgment
The obligation of a decision-maker, including an arbitrator, to make decisions based on their own evaluation rather than outsourcing reasoning to an AI tool or other external source.
Disclosure
The act of informing the other side or the tribunal about a relevant fact, process, relationship, or use of a tool. In AI disputes, this may include disclosure of material AI use.
Confidentiality
The restricted handling of information that parties want to keep private or limited to certain people.
Privilege
A legal protection that can apply to certain communications or work product, depending on the context and applicable law. Privilege is not the same as ordinary confidentiality.
Trade secret
A form of protected confidential business information that derives value from not being generally known and is subject to reasonable efforts to keep it secret.
Protective order
A procedural order designed to control how sensitive information may be disclosed, used, or accessed during a dispute.
Expert determination
A dispute process in which a technical or subject-matter expert resolves a defined issue, often more narrowly than a full arbitration or court proceeding.
Mediation
A non-binding process in which a neutral helps parties try to reach a voluntary resolution.
Litigation
Dispute resolution through court proceedings.
Seat of arbitration
The legal location of the arbitration. The seat can matter because it influences the procedural law governing the arbitration.
Governing law
The law chosen to interpret the contract or certain issues in the dispute.
Institutional rules
The procedural rules provided by an arbitral institution, such as JAMS or AAA-ICDR, that govern how a case will be administered if the parties adopt them.
JAMS AI rules
JAMS Artificial Intelligence Disputes Clause and Rules, effective June 14, 2024, designed for AI-related disputes and procedures.
AAA-ICDR AI guidance
AAA-ICDR’s March 2025 Guidance on Arbitrators’ Use of AI Tools, which addresses accuracy, fairness, independent decision-making, transparency, and confidentiality.
AAAi Standards
AAAi Standards for AI in ADR published by AAA-ICDR, offering a broader human-centered ethics and governance framework for AI use in alternative dispute resolution.
Ciarb AI guideline
The Chartered Institute of Arbitrators Guideline on the Use of AI in Arbitration, launched on September 5, 2025 and updated in September 2025, addressing party use, arbitrator use, disclosure, and model procedural language.
AI-native arbitrator
A system in which AI performs some adjudicative function. This remains highly contested and should not be confused with ordinary AI-assisted workflows for human arbitrators.
Human-centered arbitration
An approach to arbitration that uses technology, if at all, to support rather than displace human judgment, accountability, and fairness.
Conclusion
The vocabulary of AI dispute resolution is still evolving, which is one reason confusion spreads so quickly in this space. Clear terms do not solve every dispute, but they do make better thinking possible.
Use this glossary as a reference point while the Sherafy section expands.
Further Reading
- 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
- JAMS Artificial Intelligence Disputes Clause and Rules: https://www.jamsadr.com/artificial-intelligence-disputes-clause-and-rules
- AAAi Standards for AI in ADR: https://www.adr.org/news-and-insights/the-aaai-standards-for-use-of-ai-in-adr/
