|
6 May 2026 (New York, NY) - For those readers who were with me last month at the Deepfake Forensics Workshop, or at Multimedia 2026 (most sessions were focused on detection, attribution, and adversarial forensics) you may have had time to attend the sessions on what the U.S. judicial and legislative systems are doing about deepfakes.
At both of these events the chatter was about "a new phase": rapid, targeted expansion of regulation, particularly regarding sexual deepfakes and political manipulation. As of early 2026, the approach has seem to shift from inaction to a complex, multi-layered framework involving new federal criminal laws, state legislation, and increased pressure on AI platforms to implement detection and removal tools.
And then you get the following story from our media partner Law360.
According to the story, very few federal judges have handled challenges to audiovisual evidence that litigants claim has been faked by artificial intelligence, raising questions about whether changes to the rules of evidence are actually necessary.
Nearly all the almost 1,000 federal judges who responded to a Federal Judicial Center survey — listed on the agenda for tomorrow's meeting of the Judicial Conference's Advisory Committee on Evidence Rules — said they haven't yet handled so-called deepfake challenges in their courtrooms. And the few judges who have reported handling the challenges said they took different approaches, survey results show.
The committee, which is tasked with updating the federal evidentiary rules to ensure they remain fair, has been evaluating a draft Rule 707, which concerns machine-generated evidence introduced without expert testimony, and is also considering issuing a proposed Rule 901(c) to govern "potentially fabricated or altered electronic evidence."
But it's unclear how the committee will proceed, based on evidence from the survey that it commissioned in December 2025, which asked judges across all federal courts about their experiences with litigant challenges to audio, video or photographic evidence that those litigants claim was faked.
The dearth of reported deepfake challenges may mean any revisions to the rules of evidence are unnecessary. But some scholars worry that the survey results suggest that deepfakes will soon be a problem for courts, and the inconsistency in how judges are handling these challenges shows that reforms are needed.
Commenting on the potential draft rule to deal with deepfake evidence, Maura R. Grossman (professor at Osgoode Hall Law School of York University in Toronto, Canada) said:
"I'd like to see them put something out. Whether they'll decide to do that and whether the results from the FJC survey are strong enough to do that is a question".
The Federal Judicial Center declined to comment on the survey results to the Law360 reporter, or to Grossman's comment.
Note to readers: Maura R. Grossman is a prominent expert at the intersection of artificial intelligence, ethics, and the law, with a specialized focus on how deepfakes impact the justice system. Grossman is widely recognized for her work on AI-generated evidence and its potential to undermine truth in legal proceedings.
Under the proposed Rule 901(c), a litigant challenging an item as a deepfake must provide evidence supporting a finding of fabrication. The item is then inadmissible unless the party offering the item can prove, by a preponderance of the evidence, that the item is authentic.
The survey itself was conducted over two weeks in January 2026 and garnered a 45% response rate, according to the report of its results. It turns out, however, that very few federal judges have actually dealt with deepfake challenges, the results showed:
- Nearly all the 931 federal judges who responded to the survey — 98% of respondents — reported that they hadn't seen a challenge from litigants to the introduction of video, audio or photographic evidence on the basis that it was falsified by artificial intelligence, according to the report.
- Just 15 judges, nearly all of whom are district or magistrate judges, said they had encountered a deepfake challenge, the report said.
None of the scholars who spoke with Law360 seemed surprised by the results.
That's because courts are just starting to see issues with deepfakes emerge. And most of the deepfake challenges that have been raised have been brought in state court, where some, particularly self-represented litigants, have created fake recordings or text messages to gain an edge in divorce and custody battles, according to scholars.
Said Rebecca Delfino, professor at Loyola Law School, Los Angeles, who has also recommended changes to Rule 901 in light of AI deepfakes:
"State court judges are more concerned than the federal judiciary, primarily because where deepfakes, both audio and visual deepfakes, are most likely to have the greatest immediate impact would be, for example, in family law cases and other cases that are traditionally state law".
Grossman pointed to two recent state court cases as examples.
- In September, a California Superior Court judge in Alameda County threw out a lawsuit against a property management company from pro se plaintiffs who submitted videos, photographs and text exchanges that the judge found had been altered.
- In another case, the New York Court of Appeals in February reversed a lower court that had admitted into evidence video clips that appeared to show instances of child sexual abuse, finding that the video clips hadn't been properly authenticated.
The Federal Judicial Center survey results identify very few similar cases at the federal level, though.
Another explanation for federal judges' lack of experience with deepfake challenges is that federal trials are uncommon, with most evidentiary disputes being settled. And many federal audiovisual exhibits come from institutional sources such as police body cameras and surveillance systems, according to professor G. Alexander Nunn at Texas A&M University School of Law.
Scholars - and the judges themselves - disagree about what the survey results mean for possible reforms to the rules of evidence.
In addition to asking federal judges about their experience with deepfake challenges, the Federal Judicial Center surveyed judges about whether they thought the rules needed to be changed to address challenges to the authenticity of the evidence.
There was a divergence of opinions between judges who had dealt with deepfake challenges and those who hadn't, according to the results:
- Two-thirds of the judges who had experience with deepfake challenges said the rules of evidence shouldn't be changed to address challenges to potential deepfakes
- Meanwhile, 55% of the judges who hadn't yet encountered a deepfake challenge said the evidentiary rules on authenticity need to be amended
Nunn said:
"Taken together, it seems like the judges who have actually wrestled with these issues seem to be saying we can already handle this under the rules we have. It's primarily judges without that experience who seem to be saying the problem is new enough and dangerous enough that they want a rule before it becomes common".
And, many scholars point out, it's difficult to draw any conclusions from that dichotomy, though, since the sample size is so small — only 15 federal judges reported handling deepfake challenges.
But some don't think a rule change is necessary or imminent. Said Vanderbilt Law School professor Edward K. Cheng:
"The fact that there are very few judges who have come across a deepfake challenge is telling. It's not to say that that's not going to be a problem, but it does suggest that all of this discussion is, on the one hand, good because it's proactive, but, on the other hand, maybe slightly premature".
But the fact that judges who did report dealing with deepfake challenges handled those challenges very differently supports moving forward with a potential Rule 901(c), Delfino said.
More than half, 53%, of those judges, for instance, said they hadn't required the litigant challenging the evidence to make an initial showing before looking further into the falsification charges.
Of the judges who did require an initial showing, three said they required some reasonable basis or argument for the challenge, two said they required some kind of minimal evidence, and two reported that they required expert evidence.
Delfino thinks those results position the advisory committee to move forward with a version of Rule 901(c):
"I think when similarly situated judges respond differently to the same problem, to me that suggests that rule makers and policymakers should pay close attention to that".
Grossman, who would also like to see Rule 901(c) put out for public comment at the committee's May meeting, comes down in the middle on whether she thinks that will happen. She initially proposed Rule 901(c) along with former Duke University School of Law professor and federal Judge Paul W. Grimm.
The committee can take any step, including offering the proposed rule for public comment on its own or in combination with a revised Rule 707, a previous version of which has already been put out for public comment. Or, the committee could hold both rules back for another year.
Grossman said:
"That's a big question - whether they will go forward with this package for people to comment on or whether they're going to say: 'Well, we've only had a handful of examples, nobody's really jumping up and down and saying it's an emergency. So let's wait. Let's wait another year and see what happens'.
I think they could go either way".
Note to readers: this past March at Legalweek in NYC, we conducted an extensive video interview with Maura about these very issues. That interview will publish within the next week.
| 
