Evaluating unacknowledged AI-generated evidence

Who should read this?

• Judges (including magistrates, commissioners, and judicial officers) who review and rule on evidentiary issues
• Judicial support staff who assist in the processing and review of evidence
• Attorneys and litigants who submit evidence to understand the judicial expectations for evidence presentation, particularly regarding digital and AI-generated information

Why this guidance matters

While fabricated evidence is not a new problem for the courts, the increased accessibility of artificial intelligence (AI) has made it easier than ever to enhance or alter evidence, or to create fake digital evidence that looks convincingly real. Meanwhile, the increased sophistication of AI technology has made detecting such manipulation more difficult than ever. sues related to evidence that is actually or allegedly generated or altered by AI may take two forms: one where the use of AI is acknowledged, such as disclosed enhancement of a photo or audio ("acknowledged AI-generated evidence"), and a second where evidence is offered by one party and another party challenges its authenticity on grounds suggesting possible generation or alteration using AI (alleged "unacknowledged AI-generated evidence"). 

If not properly evaluated, the potential consequences of unacknowledged AI-generated evidence could lead to miscarriages of justice. This bench card addresses the issues raised by the second case, suggesting questions or areas of inquiry that may help inform a trial court's determination about
the authenticity of digital evidence alleged to be unacknowledged AI-generated evidence.

Download the bench card