Have a company transfer copyright protection to a user of the company's AI application
Have the creator of a model maintain copyright protections for their content
Have publicly available data be scalable for use in LLMs, in the same way that searchable content can be indexed. So it can be illegal to re-sell a published AI work, but it is not illegal to scrape and train on that work.
It's important that it be possible to: