There's a very good, succinct paper from the Congressional Research Service that summarizes the many interactions between generative AI systems (both in terms of the training sets used to develop these systems and the outputs that such systems produce) and copyright law. There are many, many more questions than answers here, but this is a helpful summary of the issues.

See

https://crsreports.congress.gov/product/pdf/LSB/LSB10922

Once again, I'm indebted to Gary Price for the pointer to this report.

Clifford Lynch
Director, CNI