I think copyright is a real world counter example to this kind of argument (that raw economics will override the legal protection). If you are selling anything, you have to respect copyright. So if training on copyrighted images becomes illegal, it doesn’t mean that people won’t be able to create stable diffusion models (they will do it regardless). But it does mean that businesses won’t be able to use them, which removes almost all financial benefits from creating them.

Proof both ways is difficult. How will a company prove that an image was generated with an illegally-trained model? Unless there becomes some requirement that models be kept around for a century.

They can't. They can make it easier to sue the creator of a work for infringing on existing art or characters ( https://en.wikipedia.org/wiki/Copyright_protection_for_ficti... ) used in the work.

It's impossible to reconstruct the dataset from the model.

Honestly, I`m not very optimistic about it.

1) same as with torrent trackers. you can download anything, but most of the times you can't legally use it.

2) yet. also harm/good point is a bit hypocritical. pirating of software is good? what about pirating of open source?

3) that`s could become issue for a model creator.