US evidence rules committee weighs deepfake rule and revised AI-evidence Rule 707

The US federal judiciary's Advisory Committee on Evidence Rules met in Boston on Thursday 15 October 2026 with two expert panels on its agenda: one on deepfakes and a draft Rule 901(c), and one on a revised proposal for a new Rule 707 on evidence produced by artificial intelligence. Neither rule has been adopted. This piece draws on the committee's agenda book and an IBTimes UK report of the meeting.

The deepfake draft: Rule 901(c)

The agenda says the committee "has prepared, but not finally approved" a new Rule 901(c). The working draft in the agenda book sets out two steps:

• Challenger first. "If a party challenges the authenticity of an item of evidence on the ground that it has been fabricated, in whole or in part, by generative artificial intelligence, the party must present evidence sufficient to support a finding of fabrication to warrant an inquiry by the court."
• Then the proponent. "If the opponent makes that showing, the item of evidence is admissible only if the proponent demonstrates to the court that it is more likely than not authentic."

The draft also requires "reasonable pretrial notice" from the challenger unless the court orders otherwise, and defines generative AI as "a computer system that emulates the structure and characteristics of input data in order to create derived synthetic content, including images, video, audio, text, and other digital content." The draft Committee Note justifies the higher standard on the basis that "any member of the public now has the capacity to make a deepfake, with little effort and expense, and deepfakes have become more difficult to detect by jurors."

The reporter, Professor Daniel Capra of Fordham, writes in his 15 September memo that the committee has been "unsure about the need for such an amendment." He proposes adding a list of warning signs to the Note, such as "missing or anomalous metadata", but warns of "a danger" that, "by necessarily being underinclusive, the list will lead courts to ignore or downplay other relevant factors."

Rule 901(c) has not been published for public comment. According to the minutes of the committee's 7 May meeting, the Justice Department's representative said the department "does not believe Rule 901(c) is ready for publication", and the chair, Judge Jesse Furman, noted that the rule "would not go into effect until December 1, 2028 – even if published this summer." The committee chose a fall "mini-conference" instead of a vote.

The AI-evidence draft: Rule 707

Rule 707 got further. The original "machine-generated" evidence version was published for comment in August 2025; the minutes record two hearings and 59 written comments by the 16 February 2026 deadline (the agenda's summary says more than 70). According to that agenda summary, in May the committee "voted to withdraw the proposed rule as issued for public comment, and to submit a modified rule proposal for further study." The reporter's memo says the committee "also approved in principle a new draft of Rule 707", and presents it as the version "tentatively approved by the Committee at the last meeting".

The revised working draft says that if AI-produced evidence "is offered without an expert witness, but would be subject to Rule 702 if testified to by a witness", the proponent must show it will help the trier of fact, rests on "sufficient facts or data", is "the product of reliable principles and methods", and reflects their reliable application. It adds: "Ordinarily, the proponent must provide an expert to explain how the system of artificial intelligence reliably produced the evidence. But in exceptional circumstances, the court may consider other proof to find that the output satisfies the requirements of (a)(2)-(4)." The memo adds that this "is not to say that the Committee has approved all of the language".

Who was on the panels

The agenda book lists the deepfake panel as Professor Hany Farid of Dartmouth; Professor V.S. Subrahmanian of Northwestern; James Horan of NIST; Shira Anderson, described as "Policy Counsel at OpenAI"; Devin Norder, described as "a Director of Product Management at Meta"; the digital forensics expert witness Lars Daniel; and Sean Broderick of the courts' Defender Services Office. The Rule 707 panel mixed academics, NIST and practising lawyers from both prosecution and defence.

What has been reported

IBTimes UK reported on 16 October that Ms Anderson warned the committee: "AI images are only getting better, and I do not think that there will be a scientific way going forward to detect them," and cautioned that watermarking and similar measures would not be "bulletproof". It says Devin Norder told the panel that media from recent advanced models could be indistinguishable to humans from authentic material, and that Judge Valerie Caproni, a committee member, cautioned against rushing into an AI-specific rule without stronger evidence that deepfakes were becoming a widespread courtroom problem. IBTimes UK also reports that Judge Furman appointed a subcommittee to examine the issue further. The May minutes record a Federal Judicial Center survey in which only 15 of 931 responding judges reported having encountered a deepfake challenge.

The UK position

In England and Wales, the judiciary's AI guidance for judicial office holders, dated 31 October 2025, states: "AI tools are now being used to produce fake material, including text, images and video. Courts and tribunals have always had to handle forgeries, and allegations of forgery, involving varying levels of sophistication." It asks judges to be aware of the "potential challenges posed by deepfake technology."

Why it matters

On our reading, the agenda book shows a committee with a worked-out mechanism for deepfake challenges that has not yet decided whether a new rule is needed at all. Rule 901(c) was not published this summer, and the chair's May estimate of December 2028 assumed that it would be, so on our reading any federal deepfake rule is still years away. Rule 707 is back in redrafting after public comment.