OpenAI lawyer doubts reliable AI-image detection will exist
Shira Anderson told a US Judicial Conference evidence rules committee that detection tools vary by company and will not become fully reliable.
OpenAI policy lawyer Shira Anderson said at an October 15 meeting in Boston that she does not think a scientific method for detecting AI-generated images will exist in the future, according to CNET Japan. She said detection systems differ between companies and will not become completely reliable.
The meeting was held by the US Judicial Conference's Advisory Committee on Evidence Rules to discuss deepfake evidence in court.
A draft rule under discussion would require the party submitting disputed evidence to show the court it is more likely than not authentic once the challenger offers some basis for doubting it. As of the materials for the October 15 meeting, the proposal had not received final approval.
OpenAI said some of its images carry C2PA provenance information and digital watermarks, but the company noted this can be lost through editing or file format conversion.
Judges report deepfake objections
Draft minutes from a May 7 meeting said 15 of 931 federal judges surveyed had encountered an objection that evidence was a deepfake.
Quick answers
What did OpenAI's Shira Anderson say about AI image detection?
She said she does not think a scientific method to detect AI-generated images will exist in the future, and that detection systems differ between companies and will not become completely reliable.
What is the draft court rule about deepfake evidence?
It would require the party submitting disputed evidence to show the court it is more likely than not authentic once the challenger offers some basis for doubting it. It had not received final approval as of the October 15 meeting materials.