Piracy websites should adopt Midjourney's arguments on copyright:

"Look bro, it's just too hard to figure out what license governs each and every file and who its author is."

@siguza just disguise the next filesharing service as an AI, which surfaces it's training data.
Problem solved!
@mxk instead of a download link, it gives you a string that looks like an AI prompt that you can copy-paste into a text field on the website. Any unknown prompt will just give you about 70MB of /dev/random.