Skip to content

Courts on AI / evidence / rules

Draft Federal Rule of Evidence 901(c)

Deepfake authentication burden-shifting draft · proposed rule

Status as of

Under study

“The Committee has largely coalesced on what a rule amendment might look like.” — source

Next: Potential publication for public comment was discussed for the Spring 2026 meeting; no final text was advanced in the checked drafts.

Draft Rule 901(c) would require an opponent to make a threshold showing that generative AI fabrication warrants inquiry, then require the proponent to prove authenticity by a preponderance. The May 2026 agenda treats the draft as under study pending survey results and committee action.

Operative text

Rule 901(c). Potentially Fabricated Evidence Created by Generative Artificial Intelligence. (1) Showing Required to Warrant an Inquiry into Fabrication. 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. (2) Showing Then Required by 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. (3) Applicability. This rule applies to items offered under either Rule 901 or 902. (4) Notice. Unless the court orders otherwise, a party claiming that an item is fabricated in whole or part by generative artificial intelligence must provide reasonable pretrial notice to all opposing parties of the intent to present evidence of fabrication, so that the opposing parties have a reasonable opportunity to respond to that evidence before trial. (5) Definitions. In this rule, generative artificial intelligence means 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.

Verbatim; United States government works are public domain.

Decisions that cite it

  1. Burnley v. ValentinU.S. District Court for the Eastern District of Virginia · 2026-03-18 · admitted

    Walburn offered a phone recording to enforce a settlement. Burnley alleged the voice was either someone else or an AI clone. The district court accepted sworn declarations about creation and transfer of the recording, plus voice identification under Rule 901(b)(5), and enforced the settlement with fees.

  2. State v. AmydaIowa Court of Appeals · 2026-01-28 · admitted

    The prosecution introduced a Facebook Messenger video showing sexual contact with a sleeping victim. The court upheld authentication through circumstantial evidence identifying the victim, bedroom, clothing, and the defendant’s tattooed hand, rejecting Amyda’s unsupported deepfake theory and affirming the conviction.

  3. In re D.S.Superior Court of Pennsylvania · 2024-11-14 · admitted

    The Commonwealth used an Instagram Live video allegedly showing D.S. holding a firearm. On appeal he suggested manipulation or deepfake concerns, but the Superior Court treated those arguments as waived because they were not raised at trial; authentication rested on witness recognition and platform explanation.

  4. Commonwealth v. SponeSuperior Court of Pennsylvania · 2023-11-14 · admitted

    Anonymous texts containing images and videos of minors were traced through Pinger records, IP evidence, and phone data to Spone. Some witnesses believed media had been manufactured and public reports used deepfake language, but the courts did not resolve AI generation and affirmed the convictions.

compiled by the Cognesio LLP editorial teammethodas of 2026-09-07v0.1.1coding CC BY 4.0JSON Hugging Face GitHub