OpenAI and Meta warned federal judges that tools for spotting AI-made audio and video may never be fully reliable, though only 15 of 931 surveyed judges have faced such a dispute.
Key Points:
- OpenAI and Meta told a federal evidence panel in Boston that AI-made audio and video are getting harder to authenticate.
- An OpenAI lawyer said watermarks and detection tools will never be fully reliable.
- Judges are studying a draft deepfake rule but have not approved it.
OpenAI Deepfake Warning
Representatives of both companies presented their concerns on Oct. 15 at a Boston meeting of the U.S. Judicial Conference's Advisory Committee on Evidence Rules. The panel has worked on the issue since 2023.
Shira Anderson, policy counsel at OpenAI, told the committee that AI images keep improving and that she does not expect a scientific method to detect them in the future. New laws in California and the European Union already require watermarks. But each company uses its own system, Anderson said, so no universal method can verify all visual content, and she doubted the tools would ever be "bulletproof."
Devin Norder, director of product management at Meta, said people cannot tell media made by the latest AI models apart from authentic content. Producing it now takes little time or money, he added.
Also Read: Bitcoin Holds $80,000: A Pause Before Challenging $87,000?
Jesse Furman's Deepfake Review
U.S. District Judge Jesse Furman of Manhattan, who chairs the panel, said academic predictions about deepfakes in court have not yet shown up often in real cases. The numbers back him up. In a Federal Judicial Center survey from March, only 15 of the 931 judges who responded had seen a party challenge evidence as a deepfake.
Furman appointed a subcommittee to dig deeper. Members have drafted, but not approved, an amendment called Rule 901(c), under which a party alleging fakery must first back up the claim before the other side has to show the material is more likely than not authentic.
One member urged restraint. Judge Valerie Caproni cautioned against rushing into an AI-specific rule without stronger evidence that deepfakes have become a widespread problem in courtrooms.
She argued that courts have dealt with people offering false evidence for a long time, and that existing laws already prohibit fraudulent submissions.
In May 2025, the panel declined to adopt legal experts' proposals for deepfake rules, as members argued that existing authenticity standards were adequate for AI evidence.
Months later, Alameda County Superior Court Judge Victoria Kolakowski dismissed a California housing case after finding that the plaintiffs had passed off an AI-generated witness video as real evidence.
Read Next: Crypto's 39.6% Rally Has A Problem: Spot Buyers Aren't Keeping Up
