Proposed Federal Rule Governing Admissibility of AI Evidence

In its May 6, 2026, Report, the Advisory Committee on Evidence proposed a new Rule 707 to address AI evidence.  The proposal is the product of a three-year study and is designed to address reliability and authenticity of both acknowledged and deepfake evidence.  The Advisory Committee’s proposal is:

Rule 707. Evidence Produced by Artificial Intelligence and Presented at Trial Without an Expert

(a) General Rules. If evidence is a product of artificial intelligence and is offered without an expert witness, but would be subject to Rule 702 if testified to by a witness, the proponent must establish that the evidence:

(1) will help the trier of fact;

(2) is based on sufficient facts or data;

(3) is the product of reliable principles and methods; and

(4) reflects a reliable application of the principles and methods to the  facts of the case.

(b) Establishing Admissibility. Admissibility under this rule ordinarily requires the proponent to provide an expert to explain how the system of artificial intelligence reliably produced the evidence; but in exceptional circumstances, the court may rely upon other proof to find that the output  satisfies the requirements of (a)(2)-(4).

(c) Notice. The evidence is admissible only if the proponent gives an adverse party reasonable notice of the intent to offer the evidence — so that the party  has a fair opportunity to meet it.

(d) Judicial Notice. This rule does not apply to facts that may be judicially  noticed under Rule 201.

(e) Definition. In this rule, “artificial intelligence” means a machine-based  system that can, for a given set of human-defined objectives, make  predictions, recommendations or decisions influencing real or virtual environments.

The Advisory Committee wrote that “many commentators… stated that the time is ripe and that the problem of AI being presented without expert testimony is occurring today, especially in criminal cases.” It added:

Another important issue of dispute raised by public commentary was whether AI-produced evidence should ever be admissible without an expert foundation. Some argued that an expert foundation should always be required. Others argued that an “expert-only” foundation was unnecessarily strict and that AI could be shown to be reliable through evidence such as unbiased validation tests — or perhaps that, in the future, AI could be “self-validating,” in which event an “expert-only” requirement would be outmoded.

However: “The Committee reviewed all the public comments and considered several alternative redrafts of Rule 707 prepared by the Reporter. Committee members preferred the draft that provided a presumption that an expert foundation was required but also allowed for the possibility that other evidence might show that the AI is reliable.”

The Advisory Committee wrote that a “few commentators were of the view that the time is not ripe for an amendment requiring that AI generated evidence offered without an expert must be reliable because, at this point, there is no problem to address, and the Committee should wait to see how AI develops.”

One article in the “no problem exists” camp is Cohl K. Love, What Gap? The Redundancy of Proposed Rule 707 (ABA Litigation Section Aug. 20, 2026).