§ 01 · What this is
AI disputes are a distinct category of commercial dispute with their own evidentiary character. The artifacts at issue — model weights, training data, prompts, system cards, evaluation reports, retrieval logs, inference timestamps — are not the artifacts of conventional commercial arbitration. They require a tribunal that can scope discovery to what is probative, design tiered protective orders that distinguish weights from outputs, and pose its own technical questions where the record is incomplete.
Daniel arbitrates under emerging AI dispute frameworks — JAMS AI Disputes Clause and Rules where parties have incorporated them; party-stipulated frameworks where they have not. He has authored and co-authored standing protective-order templates for AI matters.
Engaging an AI arbitrator
As an AI arbitrator, Daniel sits on disputes where the record is the model itself — weights, training data, system cards, eval reports, and inference logs. The role is distinct from conventional commercial arbitration: it requires a tribunal that can scope discovery to what is probative, design tiered protective orders, and pose its own technical questions where the record is incomplete.
§ 02 · When to retain
Indicators that this is the right role for the matter.
01
Model output liability
Hallucinated outputs in regulated workflows, defamation, autonomous-system intervention failures.
02
Training data IP
Disputes over the provenance, licensing, and lawful use of training corpora.
03
Algorithmic bias / discrimination
Statutory and common-law claims arising from differential treatment in automated decisioning.
04
AI vendor / enterprise contract
Disputes over service-level guarantees, model performance representations, and use restrictions.
05
Generative IP
Copyright, derivative-works, and substantial-similarity disputes over generative outputs.
06
Embedded ML in safety-critical products
Disputes where ML behavior contributed to a physical-world harm.
§ 03 · Methodology
How an engagement runs.
Typical scale: Six- to eighteen-month proceedings. $5M to $1B+ exposure typical.
-
01
Framework selection
AI-specific rules where the parties’ contract incorporates them. Where the contract is silent, a tailored framework can be proposed with parties’ consent.
-
02
Tiered protective order
Standing template distinguishing weights, training data, prompts, inference logs, and outputs into separate access tiers. Reviewer credentialing per tier.
-
03
Discovery scoping
System cards, eval reports, RAG configurations, retrieval logs, fine-tuning datasets, inference timestamps. What is probative; what is fishing into trade secrets dressed as discovery.
-
04
Technical adjudication
Daniel may pose tribunal questions where the record is incomplete. Hearings may include live model interrogation under controlled conditions.
-
05
Award
Reasoned award addressing both technical and legal questions. Where the matter turns on machine behavior, the award sets out the model behavior actually observed and the legal consequences turning on it.
§ 04 · Frequently asked
About this role.
Does Daniel work under specific AI dispute rules?
Where the contract incorporates them, yes. Where it does not, the parties can stipulate to a framework — including the JAMS AI Disputes Clause and Rules — or proceed under a tailored protocol.
Will the tribunal examine the model directly?
Where the parties stipulate and the protective order permits, yes. Direct interrogation under controlled conditions is sometimes determinative.
How are model weights protected during discovery?
Through a five-tier protective order with cleanroom review for the most sensitive artifacts. Daniel has authored these for many matters and adapts them to the specifics.
§ 05 · Representative work
See how this role plays out.
Begin a ai disputes — tribunal services matter