Anthropic has expressed the belief that copying protected media as an ‘intermediate step’ before creating ‘a non-infringing output’ can constitute fair use. Photo Credit: Mohamed Nohassi
Weeks after being named in a music industry copyright suit, Anthropic has responded to a request for comment on “a number of copyright issues” pertaining to generative AI, claiming, among other things, that copying copyrighted material can in certain instances “constitute fair use.”
Anthropic, which has developed products including an AI assistant, just recently forwarded its remarks to the Copyright Office. Meanwhile, multiple lawsuits have accused the developers behind leading AI systems of “training” their respective models on protected works without authorization – and, in turn, of infringing upon the copyrighted media at hand.
But as mentioned, AI research company Anthropic – which is being sued specifically by Universal Music, Concord, and ABKCO for “systematic and widespread” alleged infringement – has claimed that its aforesaid AI assistant, Claude, was trained in a way that “qualifies as a quintessentially lawful use of materials.”
Taking things a step further, Amazon- and Google-backed Anthropic also stated that a “diverse array of cases supports the proposition that [the] copying of a copyrighted work as an intermediate step to create a non-infringing output can constitute fair use.”
“For Claude, as discussed above, the training process makes copies of information and for the purposes of performing a statistical analysis of the data,” the AI business wrote. “The copying is merely an intermediate step, extracting unprotectable elements about the entire corpus of works, in order to create new outputs.
“In this way, the use of the original copyrighted work is non-expressive; that is, it is not re-using the copyrighted expression to communicate it to users. To the extent copyrighted works are used in training data, it is for analysis (of statistical relationships between words and concepts) that is unrelated to any expressive purpose of the work. This sort of transformative use has been recognized as lawful in the past and should continue to be considered lawful in this case,” continued Anthropic.
Of course, the plaintiffs in the aforementioned lawsuits would argue against (and are arguing against) the position, as would most all of the creatives and others whose life work is being used without permission or compensation to make “new outputs” en masse.
Expanding upon the point, ASCAP in its own, comparatively in-depth comments to the Copyright Office quite directly pushed back against the idea that training AI on protected media sans authorization could ever constitute fair use.
“Based on our current understanding of how generative AI models are trained and deployed, we do not believe there is any realistic scenario under which the unauthorized and non-personal use of copyrighted works to train generative AI models would constitute fair use, and therefore, consent by the copyright holders is required,” the PRO spelled out.
“The use of copyrighted materials for the development [of] generative AI models is not transformative. Each unauthorized use of the copyrighted material during the training process is done in furtherance of a commercial purpose,” ASCAP proceeded before diving into its point-by-point argument.
Over the summer, amid negotiations on the European Union’s AI Act, the IFPI, GESAC, and others called for mandatory AI training disclosures – a call that’s since been echoed by several different organizations. More recently, September brought with it the launch of Stable Audio as well as AI-powered features on platforms including Spotify.