AI-assisted evidence review

Illustration of an overwhelmed attorney reviewing many piles of paper and digital evidence.

After a new criminal case is charged, public defenders receive digital evidence through a legally mandated process called discovery. In theory, discovery supports defendants’ constitutional rights by providing their legal counsel with access to materials that may exonerate them in court. Yet in practice, discovery is only as useful as defense counsel’s ability to review and understand what has been furnished.

A recent explosion of digital evidence has sharply limited defense counsel’s ability to review this evidence, with defenders now receiving dozens of hours of body-worn camera footage, hundreds of jail calls, thousands of social media posts, and many other electronic files on a single case. This burden is especially acute at arraignment and other early hearings, when defenders must advise clients under intense time pressure, often with only a few minutes to review the evidence provided.

To help defenders handle this explosion of evidence, we are building open-source software that uses generative AI to improve public defenders’ ability to reason over large volumes of evidence early in a case and at trial. Our platform will be built for public defenders and designed to integrate into the case management systems they already use. We will thoroughly validate the system before it is used in practice, and then will evaluate its real-world impact in a randomized controlled trial.

Contributors