Federal judges meeting in Boston heard OpenAI and Meta warn on Thursday that AI-generated images and video are becoming nearly impossible to verify. The U.S. Judicial Conference’s Advisory Committee on Evidence Rules is considering whether courts need new rules for evidence that may be deepfakes, after studying the issue for three years, since 2023. OpenAI policy counsel Shira Anderson said she did not think a scientific way to detect AI images would exist going forward. Meta product management director Devin Norder said media from frontier models in the last few generations can be indistinguishable to people from authentic media, and that making convincing fakes takes trivial time and money.

OpenAI and Meta have built tools to verify visual content, while California and the European Union have passed laws requiring watermarking (a visible or embedded mark identifying content origin) to flag AI-generated material. Anderson said different companies’ tools do not communicate with each other, so no single system can catch every fake, and she did not think verification would ever be bulletproof.

The committee has found little evidence of deepfakes in courtrooms so far. A Federal Judiciary Center survey released in March polled 931 judges; only 15 had handled a case where someone challenged evidence as a deepfake. Committee chair U.S. District Judge Jesse Furman said courts had not yet seen much of the problem predicted in academic and other commentary. The gap has divided members over whether an AI-specific evidence rule is needed. Judge Valerie Caproni argued existing rules already address fake evidence, a longstanding problem.

Separately, the article cites FBI figures that AI-related scams generated more than 22,000 complaints in 2025 and nearly $893 million in losses. Wynton Hall said AI tools can use social media to build detailed profiles and imitate loved ones convincingly; he also said the tools are cheap, easy to use, and accessible at mass scale.

Federal judges meeting in Boston heard OpenAI and Meta warn on Thursday that AI-generated images and video are becoming nearly impossible to verify. The U.S. Judicial Conference’s Advisory Committee on Evidence Rules is considering whether courts need new rules for evidence that may be deepfakes, after studying the issue for three years, since 2023. OpenAI policy counsel Shira Anderson said she did not think a scientific way to detect AI images would exist going forward. Meta product management director Devin Norder said media from frontier models in the last few generations can be indistinguishable to people from authentic media, and that making convincing fakes takes trivial time and money.

OpenAI and Meta have built tools to verify visual content, while California and the European Union have passed laws requiring watermarking (a visible or embedded mark identifying content origin) to flag AI-generated material. Anderson said different companies’ tools do not communicate with each other, so no single system can catch every fake, and she did not think verification would ever be bulletproof.

The committee has found little evidence of deepfakes in courtrooms so far. A Federal Judiciary Center survey released in March polled 931 judges; only 15 had handled a case where someone challenged evidence as a deepfake. Committee chair U.S. District Judge Jesse Furman said courts had not yet seen much of the problem predicted in academic and other commentary. The gap has divided members over whether an AI-specific evidence rule is needed. Judge Valerie Caproni argued existing rules already address fake evidence, a longstanding problem.

Separately, the article cites FBI figures that AI-related scams generated more than 22,000 complaints in 2025 and nearly $893 million in losses. Wynton Hall said AI tools can use social media to build detailed profiles and imitate loved ones convincingly; he also said the tools are cheap, easy to use, and accessible at mass scale.