The global chip behemoth NVIDIA has found itself entangled in a legal wrangle concerning the origins of data employed in AI model training. The revised complaint unveils that NVIDIA, in a bid to uphold its competitive standing, took the initiative to reach out to the pirated e - book platform Anna’s Archive, with the intention of gaining access to millions of copyright - protected books. Even after being apprised of the fact that the website's repository of books was illicit, NVIDIA's top - brass still gave the green light to this endeavor and managed to acquire roughly 500TB of data.

Moreover, the complaint highlights that NVIDIA might have also made use of data from other 'shadow libraries'. What's more, it is alleged that the company provided tools to its corporate clients to aid them in procuring pirated datasets. This purportedly amounts to 'contributory infringement' and 'vicarious infringement'. Previously, NVIDIA had attempted to defend its actions by citing 'fair use'. However, with the emergence of crucial evidence, the tide seems to be turning in favor of the copyright holders in this legal battle.