Federal Rule 702 and AI expert testimony: the 2023 amendment

The 2023 amendment to Federal Rule of Evidence 702 clarified and reinforced the proponent’s burden to establish admissibility and the reliable application of the expert’s methods. For attorneys handling AI litigation, the amendment shapes how courts evaluate AI expert testimony.

What the 2023 amendment changed

The amendment addressed inconsistent application of Rule 702 after the Supreme Court's Daubert decision. Some courts permitted expert testimony to reach the jury if it was not wholly unreliable. The amendment clarifies that proponents bear the burden of demonstrating admissibility by a preponderance of the evidence, and courts must make explicit findings that rule requirements are met before admission.

The amended rule states that an expert's opinion must reflect a reliable application of principles and methods to sufficient facts or data. Adding "reliably" to the rule text makes clear that the application itself must be reliable, not just the abstract principles and methods.

The current rule text and advisory committee materials should be reviewed directly for the governing matter. A current public version of Federal Rule of Evidence 702 is available through the Legal Information Institute.

Implications for AI expert testimony

The amended rule requires courts to make affirmative findings that rule requirements are met. Proponents of AI expert testimony must demonstrate through qualifications and methodology that testimony satisfies every element of the rule.

AI systems are complex. Applying general machine learning principles to a specific system requires close attention to system specifics. An expert who explains neural networks generally but omits architecture, training data, or deployment conditions fails the reliable application requirement.

The sufficiency requirement matters when underlying system documentation is incomplete or missing. If an expert relies on unverified assumptions about a system, opposing counsel can challenge the factual basis of the opinion.

The Daubert factors applied to AI

Courts use Daubert factors to evaluate scientific expert testimony, and those factors apply to AI testimony. Factors include testability, peer review and publication, known error rates, operational standards, and general acceptance within the relevant scientific community.

For AI testimony, testability depends on methodological specificity. Experts must document tools, versions, parameters, and procedures so another expert can replicate the analysis on identical data.

Error rates matter for AI testimony. Many AI systems publish performance metrics like accuracy, precision, recall, and false positive rates. Experts evaluating AI reliability must address these metrics and explain their relevance to the case.

Challenges to AI expert testimony under rule 702

Opposing counsel can challenge AI expert testimony in several ways under the amended rule. The direct challenge targets reliable application by arguing that an expert offered general opinions about AI without addressing the specific system, dataset, or conditions in the case.

A second challenge targets factual sufficiency. If an expert relies on unsupported assumptions about an AI system, counsel can argue that the opinion lacks sufficient facts or data. This occurs frequently when system documentation is nonpublic, incomplete, or missing.

A third challenge addresses expert qualifications. Rule 702 requires qualification by knowledge, skill, experience, training, or education. For AI cases, experts need specific experience with the relevant AI architecture rather than general technology familiarity. A statistician might lack qualifications for deep learning testimony, and vice versa.

Practical implications for retaining AI experts

The amended rule changes how attorneys retain and prepare AI experts. Attorneys should retain experts early enough to analyze the specific system rather than just the general technology. Experts must document their methodology, including specific tools, versions, and procedures, to prove reliable application.

Experts must address analysis limitations. Overstating certainty invites reliability challenges. Counsel should evaluate applicable authority, procedural posture, and the evidentiary record in the governing jurisdiction.