Among the large number of policy issues that are raised by the introduction and expanded use of Artificial Intelligence (“AI”) – its potential for the creation and spreading of misinformation,2 the elimination of jobs,3 the amplification of discrimination,4 the invasion of individual privacy,5 the creation of bogus “facts6 and, in the extreme, the risk of human extinction7 – is that of the relationship between AI and copyright. This issue has two aspects. Generative Artificial Intelligence models are “trained” using large quantities of data as inputs and, in turn, produce outputs in response to questions that are posed by their users. Specifically, “[a]n AI model is a program or algorithm that relies on training data to recognize patterns and make predictions or decisions.”8 This raises two fundamental intellectual property issues.
First, who has the rights to the data that are used as inputs for training an AI model? Specifically, do the entities that develop AI models incur liability to the owners of the copyrighted materials that are used in their training? Second, are there circumstances in which the entities that employ AI models incur liability to any of those copyright owners? That is, can the outputs that are produced through the interaction of AI models and the queries that are posed to them be found to infringe on the copyrights of the owners of the materials that have been used in their training?9