Copyright and Artificial intelligence: A Tale of Exceptional Value
It is crucial that, as the US government considers some of the legal implications of copyright for the deployment and development of artificial intelligence, we first take a step back and ensure our goals are well-defined and grounded in a clear understanding of how copyright relates to AI and an accurate observation of legal developments in other countries. Many observers have simplified the way in which different countries have dealt with copyright and AI. All those who have done this have rejected the idea that copyright was not involved, and developed legal norms that carefully limit the scope for any exceptions, with an eye to facilitating licensing.
I wrote about the updated Copyright Directive of the EU. Here, it is important to note that, despite the claims made about Japan’s laws, their provisions, as manifested by the 2018 amendments are designed in order to avoid conflict with legitimate copyright interests. Although I do not necessarily agree with Japan’s approach, I think it’s important to note that, even the exceptions that are made, according to my understanding, recognize that data mining and machine learning is in fact a violation of copyright, apply only to material that has been acquired legally, require that each work be \”minor\” compared to the TDM effort, and stipulate that license terms must also be respected. It is unclear to me whether Japan has achieved its goal of respecting the copyright required by international law. However, it’s important to know that claims that Japan no longer considers copyright an issue in AI development are incorrect.