Overview
The series explains when an AI output becomes an expert-like inference, why the offering party ordinarily must establish reliability through an expert, and what four Rule 702-based questions and evidence package support that showing. It separates method reliability from authenticity and the working Rule 901(c) deepfake proposal, while preserving the May 2026 status: both frameworks were still under study.
Learning outcomes
- Under the revised proposal, the party offering an expert-like AI inference carries the reliability showing.
- Proposed Rule 707 targets expert-like inference, not every machine measurement or retrieval.
- Reliability turns on helpfulness, sufficient data, reliable methods, and reliable case-specific application.
- An expert is ordinarily expected; convincing independent validation may exceptionally substitute, but the proposal remains unsettled.
- Authenticity is separate; working Rule 901(c) would shift a higher burden only after an opponent’s threshold showing.
Keywords
Preview every chapter · Unlock full text, audio, and video from chapter 4
PlansFAQ
Who must prove an AI inference is reliable under Proposed Rule 707?
Under the revised proposal, the party offering an expert-like AI inference must establish its reliability, ordinarily through an expert. The May 17, 2026 report describes a proposal still under study, not enacted law.
When does Proposed Rule 707 apply to computer output?
It applies when the output is an expert-like inference that would face Rule 702 if a human offered the same conclusion. Routine measurement, retrieval, or calculation may raise different accuracy and authenticity questions.
What must a proponent show for AI reliability?
The proponent must show helpfulness, sufficient facts or data, reliable principles and methods, and reliable application to the case. Validation, records, auditability, and explanation can support that showing.
How does AI authenticity differ from inference reliability?
Authenticity asks whether the item is genuine; reliability asks whether the method supports the AI conclusion. Working Rule 901(c) would raise the authenticity burden only after an opponent’s threshold fabrication showing.