Deepfakes mean 'everyone will have to expert up' for employment and divorce cases, panelists say
How can lawyers prove evidence is authentic amid the rise of deepfakes created or enhanced by artificial intelligence? That was the focus of an afternoon session at the ABA Techshow on Thursday.
“In the past, it was pretty simple,” said Stephen Embry, a former defense lawyer and the publisher of the TechLaw Crossroads blog. “We had to show the thing is what it purports to be, and oftentimes that was accepted. But that’s no longer going to be the case.”
Embry, who was joined by U.S. District Judge Xavier Rodriguez of the Western District of Texas for the panel titled, “Deepfakes in Court: When Seeing isn’t Believing,” also noted this could create real issues for judges in the future.
Evidence was traditionally created by humans, but now courts will grapple with evidence generated by artificial intelligence, which uses algorithms to synthesize and produce new content, Rodriguez said. He and Embry highlighted certain practice areas, including family law, where the introduction of AI-generated evidence could cause real harm.
“You have a domestic assault case, and the husband comes in with an audiotape of him being threatened or vice versa, and what’s the judge’s reaction?” said Embry, who is also a member of the 2026 Techshow board. “That it’s probably valid.”
Follow along with the ABA Journal’s coverage of the ABA Techshow 2026 here.
Plaintiffs or managers could also introduce deepfake statements supposedly made by the other side in employment law cases, Rodriguez added. These statements could claim to be evidence that someone made racially derogatory remarks or be used to justify a termination.
“That’s the other area I see that’s real ripe for this,” Rodriguez said.
Embry and Rodriguez discussed whether lawyers have an obligation to question evidence produced by their clients. As one example of potential guidance, Rodriguez pointed to proposed Rule 707 of the Federal Rules of Evidence. The proposed rule, which would regulate machine-generated evidence, was published for public comment in August.
Other unanswered questions include whether a judge or jury will decide if potentially problematic evidence should be admitted and what types of experts will be asked to prove the difference between real evidence and deepfakes, Rodriguez and Embry also noted.
“We will now have a system where everybody will have to expert up,” Rodriguez cautioned. “Everybody will have to get their respective forensics teams in there. For your average employment discrimination case, much less family law in state courts, I don’t know how you’re going to justify those kinds of costs.”
Write a letter to the editor, share a story tip or update, or report an error.
