Over 30 lawsuits challenging the use of copyrighted materials in training Generative AI models are currently pending, with most cases being heard in federal courts across the U.S. These copyrighted materials include news articles, photographs, and music. The legal landscape remains unclear as to whether such training violates copyright law.
However, the uncertainty surrounding this issue means that AI training can continue until there is final guidance from the courts (or potentially from the legislature), both of which could take considerable time. A relevant example is Thomson Reuters, which provides Westlaw, suing a competitor for copyright infringement in May 2020. Last month, the Court partially granted summary judgment, ruling that the headnotes and numbering had been copied. However, the case is still pending trial.
Of course, even if a court determines that training Generative AI on copyrighted materials does not qualify as “fair use” or fall within an exception for data mining, “it will be impossible to ‘untrain’ the GenAI engines,” which are already in active use. This leaves the courts with the challenging task of deciding what remedy, if any, is appropriate once the legality of such training is resolved. Our IP lawyers are keeping a close eye on these matters, and are here to help with any IP and AI concerns.