OpenAI and Meta told federal rulemakers deepfakes may soon be undetectable. The committee kept studying. Any new federal rule is still years out.
No, there’s no new federal deepfake evidence rule, and there won’t be one soon. The federal judiciary’s Advisory Committee on Evidence Rules met in Boston on October 15, 2026 with a draft Rule 901(c) on deepfakes and a revised Rule 707 on AI-generated evidence on its agenda, heard from experts including an OpenAI lawyer who doubts detection will ever keep up, and adopted neither; the committee agenda book describes the meeting as expert panel discussions on both drafts rather than a vote. Any new federal rule is still years out.
What Did OpenAI and Meta Tell the Committee?
The most striking testimony came from the deepfake panel. Shira Anderson, policy counsel at OpenAI, told the committee, as reported by IBTimes UK:
“AI images are only getting better, and I do not think that there will be a scientific way going forward to detect them.”
Devin Norder, a director of product management at Meta, said media from recent AI models cannot be told apart from authentic content by humans and is now cheap and quick to create, according to Investing.com’s report. Anderson also said each company uses its own watermarking system, so there is no universal way to detect every watermark.
So far, though, the evidence of a courtroom problem is thin. A Federal Judicial Center survey found that only 15 of the 931 judges who responded had seen a party challenge evidence as a deepfake. Committee chair Judge Jesse Furman noted that academic predictions have not yet shown up in many real cases. IBTimes UK reports that Furman appointed a subcommittee to keep studying the issue, and that Judge Valerie Caproni cautioned against rushing into an AI-specific rule without stronger evidence of a widespread problem.
What Would the Draft Rules Change?
Draft Rule 901(c) works in two steps. A party claiming that evidence was fabricated by generative AI must first present evidence sufficient to support that finding. If it does, the party offering the evidence must show the court it is more likely than not authentic. The agenda book says the committee has prepared the draft but not finally approved it, and it has not been published for public comment.
Rule 707 has gone further. The committee published a version for comment in 2025 that would apply expert reliability standards to machine-generated evidence offered without an expert. After two hearings and dozens of written comments, the committee voted at its May 2026 meeting to withdraw that version and study a revised one. The May minutes in the agenda book show members preferred a middle option: admissibility would ordinarily require an expert to establish reliability, but in exceptional circumstances the court could rely on other proof, such as validation studies.
How long could this take? The May minutes record Furman saying Rule 901(c) would not take effect until December 1, 2028, even if it had been published in summer 2026. It was not.
So Who Carries the Deepfake Burden in Court Right Now?
Trial teams do. Nothing changes in federal court today: existing Rule 901 already requires the party offering evidence to show it is what it claims to be, and challenges to recordings and images proceed under that rule. The October meeting shows that the rule writers take the problem seriously and have a worked-out mechanism ready, but are not yet convinced it is needed.
If the people building image models don’t think detection will keep up, we wouldn’t bet a case on a detection tool alone. Provenance, chain of custody and metadata are your friends, and your forensic experts will carry more of the load.
How Should Litigation Teams Prepare?
At a large litigation department or firm, it is worth checking that your e-discovery and forensic vendors can document where audio, video and images came from, beyond whether they look altered. Ask what the vendor would testify to.
If your practice is employment, family or insurance work, where recordings and photos are common evidence, know which forensic expert you would call before the first challenge arrives. Small firms and solo litigators have a reason to keep an eye on state courts as well: Furman noted at the May meeting that the federal rules are a model for state courts and that AI issues may surface first in places like family court.
The next real signal will be whether the committee publishes Rule 901(c) and a revised Rule 707 for public comment, possibly together. Until it does, the old rule and good evidence hygiene are what you have.
FAQ
Did the federal rules committee adopt an AI evidence rule?
No. At its October 15, 2026 meeting it held expert panel discussions on both drafts, not a vote.
Which AI rules are under consideration?
A draft Rule 901(c) on deepfakes and a revised Rule 707 on AI-generated evidence.
Who warned the committee about deepfakes?
Shira Anderson, policy counsel at OpenAI, and Devin Norder, a director of product management at Meta, who said recent AI media cannot be told apart from authentic content by humans.
When could a new rule take effect?
Not soon. Any new federal rule is still years out.
*Updated Oct 23 to refresh the story and add an FAQ.*
In this story
See every round in the Funding Tracker