What the revised text will say is not public. The committee has not released its changes, and it has not set a date for the rule's return. Lawyers' groups and a federal courts committee backed pretrial reliability screening but asked for narrower language, such as "machine-generated inferential evidence." Critics attacked the phrase "machine-generated evidence" as broad enough to reach geolocation records and surveillance footage. They also objected to a vague "simple scientific instruments" exception and to the risk that AI output could enter without an expert. Under the current timeline, the rule cannot take effect before December 1, 2029. One commentator estimates republication will not occur until August 2027 at the earliest, which is a projection, not a schedule. Regulators have not abandoned the rule.
For attorneys, the next checkpoint is the October 2026 advisory committee meeting, where the revised draft could return. Until then, no federal standard governs AI-generated evidence, and the rule would not reach deepfakes in any case, because it covers only evidence a party admits was AI-made. A separate deepfake-authentication amendment to Rule 901 was set aside in favor of Rule 707. Judges and lawyers will keep handling AI-generated evidence under existing rules and local practice, so challenges to authenticity and reliability will continue to rely on current law.