This is a slippery slope. ChatGPT, in most cases, won’t be found to have “stolen” the work but rather “learned from” the work. It is similar to how we learn – training the neurons and synapses to fire in certain patterns based on years of training.
I do think attribution should be required when a certain threshold of reliance on an individual piece of work exists but my guess is that most of the responses generated by ChatGPT are going to rely on training from 1000’s of sources with minimal amounts being attributed to one piece of work.
