Copyright and AI

LLMs have been trained on most of the internet and sometimes the model developers have not had the right permissions for the training data they have used.

Many content creators (including song writers, photographers and authors) argue that AI LLM developers have “breached” copyright and stolen their material without permission.

A recent court case ended with Anthropic (developers of the Claude LLM models) paying authors for pirating books used in their training. But that does not mean that LLM developers can't use books in their training!