OpenAI and Meta warn US judiciary about AI evidence

Serdar HocamAuthor & Editor

AI giants have reported that advancing deepfake technologies will make it difficult to verify audio and video evidence in future court hearings.

◉ 3 views
OpenAI, Meta caution US judiciary about risk of 'deepfake' evidence

Representatives from OpenAI and Meta stated that AI-generated deepfake content has advanced to the point where it is increasingly difficult to determine the authenticity of audio and video evidence presented in courts.

Critical Meeting in Boston

Technology experts from OpenAI and Meta shared their views by attending a meeting of the Advisory Committee on Evidence Rules of the US Judicial Conference in Boston. The committee continues to examine the risks of AI-generated fake evidence in courts.

Risk of Inability to Detect

OpenAI policy advisor Shira Anderson conveyed to the committee that AI visuals are continuously getting better and that in the future there may not be a scientific way to detect them. The committee has been grappling with the issue of how deepfake evidence will meet reliability standards since 2023.

Cases Faced by Judges

A survey conducted by the Federal Judicial Center in March showed that only 15 out of 931 participating judges had a party to a case object to evidence as a deepfake. However, experts emphasized that this issue is not entirely hypothetical.

Meta's Emphasis on Cost and Realism

Meta's director of product management, Devin Norder, stated that media produced by frontier models cannot be distinguished by humans from authentic media, and that the cost and time required to produce such content are very low.

Watermark and Detection Limitations

While it is known that companies use verification tools and that there are watermark laws in California and the European Union, Anderson noted that there is no universal way to reliably detect all watermarks.

Cautious Approach on the Judicial Side

US District Judge Valerie Caproni spoke cautiously, stating that there is not enough data showing the problem is widespread and that there should be no rush to create a rule specific to artificial intelligence. Caproni reminded that existing rules already make presenting fake evidence illegal.