Federal Judicial Center Survey Finds Few Deepfake Challenges in U.S. Courts

Federal prosecutors in the Western District of Texas indicted an innocent person on fabricated audiovisual evidence last winter. The material, produced by a confidential informant, supported the indictment against a person whose case the government later dismissed without explanation.

The fabrication came to light only when the informant pleaded guilty to obstruction of justice at his own sentencing hearing.

Nobody challenged the audiovisual file as a deepfake during the original prosecution. Counsel for the dismissed defendant did not have the file in hand to raise an objection at the time it mattered. The most damaging deepfake in the survey reached federal court through a guilty plea.

The case appears in the first federal survey of judges on how courts handle deepfake challenges. The survey was released on March 25, 2026 by the Federal Judicial Center. Forbes reported the details in an article published on May 14, 2026.

Of the 931 federal judges and magistrates who responded to the survey, the response rate was 45 percent. Only 15 of the 931 had ever fielded a challenge to audiovisual evidence as a deepfake. Two-thirds of those 15 judges had seen just one such challenge across calendar years 2024 and 2025.

Most deepfake challenge cases in the survey were civil. The Federal Judicial Center pulled four of the challenged deepfake cases from PACER. One was tossed on a procedural deadline before the court reached the merits.

Another of the four pulled PACER cases involved a zoomed-in clip of evidence already admitted. In the Western District of Texas case, a confidential informant produced fabricated audiovisual material to federal prosecutors.

Of the 914 judges who had not encountered a deepfake challenge, 745 said they would demand an initial showing before inquiring into authenticity. That equals 82 percent. The 15 judges with first-hand deepfake challenge experience were closer to evenly split on demanding an initial showing.

When pressed on what the initial showing for a deepfake challenge should look like, 679 judges wrote in answers. Forty-two percent of the 679 judges who wrote in would accept minimal evidence such as an affidavit, technical work pointing to inauthenticity, or comparable corroboration.

Another 35 percent would settle for a reasonable argument or a good-faith basis for asserting falsification beyond speculation.

The survey shows what readiness looks like from the bench after federal and state judges told reporters late last year they were not ready for AI-generated evidence in court. The Texas indictment demonstrates why the count of past challenges does not measure the size of the problem. Some of the most damaging fakes get through because nobody on the defense side knew to challenge them.

Proposed Rule 901(c) would build on the existing framework of Federal Rule of Evidence 901. The new wording would require a challenger to put forward evidence sufficient to support a finding of fabrication before the court probes further. A bare assertion would not clear the bar.

Federal judges by and large arrived at that position without waiting for the rule. The judges sorted the kinds of attacks they would credit along two axes. The first is the level of showing the proof the challenger has to bring, which can be an affidavit, a technical analysis or a corroborating fact.

The second axis is the nature of the attack. A direct attack points to something in the file itself such as a glitch in the video. An indirect attack points outside the file such as an alibi where the depicted location, person or event can be falsified by external fact.

Both axes line up with digital forensic work. Reading a file for direct tells is one job. Reading the device that supposedly produced or stored the file is a separate one that pulls in different records including timestamps that conflict with the claimed chain of custody.

Generative AI detection tools alone will struggle here because detection scores a file but does not authenticate the file. A probability score from a detection model can support an initial showing but cannot finish the job. The proof that survives a federal trial runs through the file, the device and the records that surround the device.

For litigants raising a deepfake challenge the survey is a road map to bring an affidavit, technical analysis, a specific anomaly in the file or an external fact the file cannot account for. For the offering side the survey points at what is coming as authentic files will still draw challenges.

Most will fail under the bar judges describe while the ones that succeed will force authentication work the proponent has to budget for.