In a report for the Foundation for American Innovation, Anupam Chander considers regulations governing AI training in the US, EU, Japan, and China in order to ascertain whether training LLMs on unlicensed works violates copyright. Although the four regions all recognize an exception to copyright for AI training, they differ with respect to how to enact this exception. Chander also considers cross-national implications: "Could a company legally train an AI system on unlicensed copyrighted works where such training is permitted, and then deploy the AI system in another jurisdiction--without incurring liability there—even if the second jurisdiction would have imposed copyright liability if the training had occurred within its borders?"
You might also like
- communityThe Cookout
- communityHow to build AI for democracy
