I recently came across an interesting aside by Bankruptcy Judge Peter Russin in the case In re: Vital Pharmaceutical (released on June 16). In the preparation of the introduction for this Memorandum Opinion, the Court requested ChatGPT, an AI language model, to compose an essay on the evolution of social media and its impact on creating personas and marketing products. Alongside the essay, ChatGPT included a disclaimer stating that it did not have access to the specific sources used in generating the essay, as it relied solely on the knowledge stored in its database. However, it mentioned that it could offer some general sources related to the topic, providing a list of five such sources. Interestingly, upon further investigation, it was discovered that none of these five sources actually existed.

Some of the sources had real authors, while others were attributed to legitimate journals. Nevertheless, all five citations appeared to be fabricated, which prompted the Court to disregard the information and conduct its own research using more traditional means. While the Court did not resort to physical books, this situation serves as a cautionary tale about the ethical dangers associated with relying on AI in its current state of development. This should be a familiar reminder, but it is worth noting again. (Coincidentally, I was recently testing Claude 2 by asking it for cases on pseudonymity in libel litigation, and it generated some fictional cases for me, similar to what has been observed with other AI programs. ) To sign up for the Volokh Conspiracy Daily email, please click here.