There are a few stages involved in delivering the output of a LLM or text-to-image model: 1. acquire training data 2. train on training data 3. run inference

There are a few stages involved in delivering the output of a LLM or text-to-image model: 1. acquire training data 2. train on training data 3. run inference on trained model 4. deliver outputs of inference One can subdivide the above however one likes. My understanding is that most lawsuits are targeting 4. deliver outputs of inference. This is presumably because it has the best chance of resulting in a verdict favorable to the plaintiff. The issue of whether or not it's legal to train on training data to which one does not hold copyright is probably moot - businesses don't care too much about what you do unless you're making money off it.