While these systems are not without risks, such as the creation of deepfakes and the facilitation of the spread of misinformation, these are, by and large, problems outside of the realm of copyright law and can be best addressed by the adoption of best practices, private ordering, and other regulations. [...] But even assuming that the use of copyrighted works as training data is considered commercial, these uses are sufficiently transformative so as to overcome the commercial nature of the entities that use them.38 In factor one analysis, it is important to weigh any commerciality against both the transformative nature of the use and the degree to which these tools “serve[] the interests of the public. [...] In the district court hearing, the court noted that, even though some of the works in question were creative and thus close to the core of copyright, “this factor is of lesser import because the allegedly infringing use makes no use of any creative aspect of the student works.”42 Similarly, in the case of generative AI, the copyrighted works are used to extract non-expressive elements despite the. [...] Fourth, the effect of the use on the market for the copyrighted works used to train generative AI systems is unlikely to be significant based on the lack of a substitutional effect between the individual works themselves and the generative AI systems based on AI models that use them as training materials. [...] The fair use statute says that the inquiry should focus on “the effect of the use upon the potential market for or value of the copyrighted work,”49 not about the effects of the byproducts of that use—in this context, the dissemination of outputs of generative AI systems.
- Pages
- 20
- Published in
- United States of America
