Beyond Regurgitation: Transformative Fair Use In The Training Of AI

AI and Fair Use

When the New York Times sued OpenAI, the allegation grabbing the most headlines concerned ChatGPT sometimes producing, as output to the user, all or substantially all of an actual article previously published by the Times. It has been widely reported that OpenAI accuses the Times of deliberately instigating GPT to "regurgitate" Times articles by "manipulating" prompts in bad faith, if not actually directing GPT to infringe the Times' own copyrights.

The wholesale unauthorized copying and distribution of copyrighted literary works to your own paying users is obviously an infringing use, every time. This is as true in 2024 with AI as it was nearly 200 years earlier, when the accomplished U.S. Representative from Massachusetts, the Reverend Charles Wentworth Upton, included 353 pages of previously published letters written by George Washington in the Reverend's own nearly 900-page, two-volume work, leading to him lose what is widely considered the fi