The 2023 amendment to Federal Rule of Evidence 702 clarified and in some respects heightened the standard for admissibility of expert testimony. For attorneys handling AI-related litigation, the amendment has direct implications for how AI expert testimony is offered, challenged, and evaluated by courts.
What the 2023 Amendment Changed
The amendment addressed a problem that had developed in the application of Rule 702 since the Supreme Court's Daubert decision: courts were inconsistently applying the gatekeeping standard, and in some circuits the rule was being interpreted to allow expert testimony to go to the jury so long as it was not wholly unreliable. The amendment clarified that the proponent of expert testimony bears the burden of demonstrating admissibility by a preponderance of the evidence, and that the court must make a finding that the requirements of the rule are met before the testimony is admitted.
The amended rule now explicitly states that the expert's opinion must reflect a reliable application of reliable principles and methods to sufficient facts or data. The word "reliably" was added to the requirement that the expert's opinion reflect an application of the principles and methods, making clear that the application itself, not just the principles and methods in the abstract, must be reliable.
Federal Rule of Evidence 702 (as amended December 1, 2023)
"A witness who is qualified as an expert by knowledge, skill, experience, training, or education may testify in the form of an opinion or otherwise if the proponent demonstrates to the court that it is more likely than not that: (a) the expert's scientific, technical, or other specialized knowledge will help the trier of fact to understand the evidence or to determine a fact in issue; (b) the testimony is based on sufficient facts or data; (c) the testimony is the product of reliable principles and methods; and (d) the expert's opinion reflects a reliable application of the principles and methods to the facts of the case."
Implications for AI Expert Testimony
The amended rule has several specific implications for AI expert testimony. First, the preponderance standard means that courts must affirmatively find that the requirements of the rule are met, not merely that the testimony clears some minimum threshold of reliability. For AI expert testimony, this means that the proponent must be prepared to demonstrate, through the expert's qualifications and methodology, that the testimony meets each element of the rule.
Second, the requirement that the application of principles and methods be reliable is particularly significant for AI-related expert testimony. AI systems are complex, and the application of general principles of machine learning to a specific system in a specific case requires careful attention to the particulars of that system. An expert who can articulate general principles of how neural networks work but cannot address the specific architecture, training data, and deployment conditions of the system at issue may not satisfy the reliable application requirement.
Third, the sufficiency of facts or data requirement has implications for AI cases where the underlying system documentation is incomplete or unavailable. If an expert's opinion depends on assumptions about the system that cannot be verified from available documentation, the sufficiency of the factual basis for the opinion may be challenged.
The Daubert Factors Applied to AI
The Daubert factors, which courts use to evaluate the reliability of scientific expert testimony, apply to AI expert testimony as they do to other technical testimony. The factors include whether the theory or technique can be tested, whether it has been subjected to peer review and publication, the known or potential error rate, the existence of standards controlling the technique's operation, and whether the technique has been generally accepted in the relevant scientific community.
Applied to AI expert testimony, these factors raise specific questions. Whether the analytical technique can be tested depends on whether the expert's methodology is sufficiently specified to allow independent replication. For AI systems, this requires that the expert document the tools, versions, parameters, and procedures used in the analysis with enough specificity that another expert could repeat the analysis on the same data and reach the same conclusions.
The error rate factor is particularly relevant for AI-related testimony. Many AI systems have documented performance metrics, including accuracy, precision, recall, and false positive rates. An expert testifying about the reliability of an AI system should be prepared to address these metrics and explain their significance for the specific application at issue.
Challenges to AI Expert Testimony Under Rule 702
Opposing counsel challenging AI expert testimony under the amended rule has several available arguments. The most direct challenge is to the reliability of the application: arguing that the expert has not adequately addressed the specific system, dataset, or conditions at issue and has instead offered general opinions about AI systems that do not reliably apply to the facts of the case.
A second line of challenge is to the sufficiency of the factual basis. If the expert's opinion depends on assumptions about the AI system that are not supported by available documentation, opposing counsel can argue that the opinion is not based on sufficient facts or data. This is particularly relevant in cases where the AI system's documentation is incomplete, proprietary, or unavailable.
A third challenge is to the expert's qualifications. The rule requires that the expert be qualified by knowledge, skill, experience, training, or education. For AI expert testimony, this means that the expert must have specific expertise in the type of AI system at issue, not merely general familiarity with AI technology. An expert with a background in classical statistics may not be qualified to testify about deep learning systems, and vice versa.
Practical Implications for Retaining AI Experts
The amended rule has practical implications for how attorneys should approach retaining and preparing AI expert witnesses. Experts should be retained with sufficient time to conduct a thorough analysis of the specific system at issue, not just the general technology area. The expert's methodology should be documented in detail, including the specific tools, versions, and procedures used, to support a showing of reliable application.
Experts should also be prepared to address the limitations of their analysis. Under the amended rule, an expert who overstates the certainty of their conclusions is more vulnerable to a Daubert challenge than one who clearly articulates what the analysis can and cannot establish. Courts have shown increasing willingness to exclude expert testimony that conflates the general reliability of a technique with the specific reliability of its application in a particular case.
AI Expert Witness Services provides expert testimony that is designed from the outset to meet the requirements of Federal Rule of Evidence 702 and the Daubert standard.
How We Work