Yesterday, the Committee For Justice filed comments in response to the U.S. Copyright Office’s inquiry concerning AI and copyright. We recommended that, rather than reinventing the wheel, the approach to copyright protection for the training material used to develop generative AI (ChatGPT, DALL·E, etc.) be modeled on how the law treats human learning.
Specifically, CFJ's comments suggest that 1) the use of copyrighted material to train a generative AI model should be presumed to be fair use, and 2) protection for copyright holders should be focused on the outputs of generative AI models by asking whether the outputs are substantially similar to copyrighted material, much as the law does for the outputs of human writers, artists, and the like, regardless of their learning process.
