Executive Summary

Copyright law depends on determinations of whether informationis “inside” an instrumentality, such as a book, hard drive, ordatabase. Modern artificial intelligence systems challenge this notionof “inside,” because the models that underlie those systems arenot simple information storage systems, but rather mathematicalfunctions representing approximations of generalized patterns. ThisArticle considers the meaning of “inside” when applied to AI models,shows that the literal assumptions of copyright law lead to intuitivelyabsurd results, and proposes new doctrinal questions for courts toanswer to better adapt the law to the nature of the technology.