The Iowa Supreme Court has held that merely suggesting a video could be an AI-generated deepfake is not enough to keep it out of evidence. In State v. Amyda, the court affirmed a criminal conviction based in part on a video depicting sexual abuse, rejecting the defendant’s argument that the recording lacked proper authentication and could have been fabricated using artificial intelligence.

The case arose after a woman received a video showing someone sexually abusing her while she was asleep. She recognized her bedroom, clothing, physical characteristics, and the distinctive tattoo on defendant’s hand. She identified defendant as the person depicted in the recording. Police investigated but were unable to recover the device used to make the video or establish its digital history. Defendant was charged with third-degree sexual abuse.

Before trial, defendant sought to exclude the video, arguing that the prosecution could not authenticate it without testimony from someone who witnessed the recorded events or could explain how the recording was made. He also invoked the best-evidence rule, maintaining that the video was a copy whose authenticity was questionable because modern AI technology can produce convincing fabrications. The trial court rejected these arguments and admitted the video. A jury convicted defendant, and the Iowa Court of Appeals affirmed.

The Iowa Supreme Court likewise affirmed. It held that Iowa Rule of Evidence 5.901 permits authentication through circumstantial evidence, including the distinctive characteristics of a recording and the surrounding circumstances. The victim’s detailed testimony identifying the room, her body, her clothing, and defendant’s hand provided sufficient evidence to support a finding that the video depicted what the prosecution claimed it depicted. Authentication did not require testimony from someone who personally witnessed the abuse or operated the recording device.

The court also rejected defendant’s argument under the best-evidence rule. Iowa Rule of Evidence 5.1003 generally permits duplicates unless a genuine question exists about the original’s authenticity. Although defendant’s attorney pointed to widely publicized examples of deepfake technology, he presented no evidence suggesting that this particular video had been manipulated. The court acknowledged the growing risks associated with convincing AI fabrications but concluded that speculation alone did not raise a genuine authenticity question.

The decision illustrates another dimension of the challenges AI presents for courts evaluating evidence. In our recent discussion of State v. Horcasitas, an Arizona appellate court vacated a criminal sentence after the sentencing judge relied on an AI-generated depiction of a deceased victim. And in our earlier discussion of Malia LLC v. State Farm, a federal court considered whether machine-generated information constituted hearsay. These cases raise different evidentiary questions, but collectively demonstrate how established rules must be applied to circumstances created by AI technology.

The significance of Amyda is that the existence of deepfake technology does not, by itself, undermine otherwise sufficient evidence of authenticity. Courts must remain alert to the possibility of fabricated recordings without allowing unsupported allegations of AI manipulation to discredit genuine evidence.

State v. Amyda, No. 24-1563 (Iowa Oct. 9, 2026)