Jeffrey Bellin (Vanderbilt University – Vanderbilt Law School) has posted Evidence Law’s AI Problem and the Hearsay Solution (Harvard Law Review, forthcoming) on SSRN. Here is the abstract:
As artificial intelligence (A.I.) seeps into our daily lives, A.I. generated evidence will increasingly appear in court. For the most part, American evidence rules are up to the task. The rules require testifying experts to demonstrate the reliability of any A.I. tools they rely on, and allow courts to deter and detect forgeries, even those generated by A.I.
There is a gap in the evidence rules, however, when non-expert witnesses testify about the outputs of artificial intelligence. For example, no evidence rule squarely prevents a victim or police officer from testifying that ChatGPT concluded that the defendant was the perpetrator. Evidence policymakers are scrambling to fill this gap. States have been unable to reach consensus on a solution, however, and a proposed federal rule that would have routed “machine-generated” evidence through the expert rules did not survive its public comment period.
This article proposes a different approach. Evidence law’s A.I. problem is that non-expert witnesses can testify about what A.I. “told” them without triggering any clear evidentiary obstacle. The evidence rules solve this problem with respect to the most complex machines of all—other persons—through the hearsay framework. Witnesses generally cannot testify about what another “person” told them; that is hearsay. Expanding the hearsay definition to also capture statements by artificial intelligence closes the gap in the evidence rules. Importantly, this does not make all A.I. statements inadmissible. Capturing A.I. statements within the hearsay definition enlists a robust, and surprisingly well-suited, framework for distinguishing A.I. evidence that should be admitted from evidence that should be excluded.
Highly Recommended!
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.
Lawrence Solum