Palmiotto on Detecting Deep Fake Evidence with AI

Francesca Palmiotto (Hertie School) has posted Detecting Deep Fake Evidence with Artificial Intelligence: A Critical Look from a Criminal Law Perspective on SSRN. Here is the abstract:

 

The widespread use of deep fakes is particularly worrisome for criminal justice, as it risks eroding trust in video, image or audio evidence. Since detecting deep fakes is a challenging task for humans, several projects are currently investigating the use of AI-based methods to identify manipulated content. This paper critically assesses the use of “deep fake detectors” from the perspective of criminal evidence and procedural law. It contends that whilst the use of AI detectors is beneficial (if not inevitable), risks arising from their use in criminal proceedings must not be underestimated. After a brief introduction to deep fake technology and detection methods, the paper analyses three key issues, namely accuracy, scientific validity and fair access to detection tools between parties. In light of these challenges, the paper argues that the introduction of deep fake detectors in criminal trials must comply with the same standards required for expert evidence. To do so, higher transparency and increased collaboration with software providers are needed.