The Copyright Office began accepting comments on August 30th, focusing on the use of copyrighted data for AI model training, the copyrightability of AI-generated material without human involvement, and AI copyright liability. Apple, however, focused its response on the copyrighting of AI-written code, arguing that such code should be copyrightable if a human developer controls its expressive elements and decisions to modify or enhance it.
Key takeaways:
- The US Copyright Office is considering new rules around AI's use of copyrighted materials and has opened a public comment period.
- Major AI companies like Meta, Google, Microsoft, Adobe, Hugging Face, StabilityAI, and Anthropic have argued against paying to train AI models on copyrighted work.
- These companies argue that AI training is similar to reading a book, that changing copyright laws could hurt small AI developers, and that the current law is adequate and should not be changed.
- Apple has focused on the copyrighting of AI-written code, arguing that it should be allowed to copyright its AI-made code.