Most pirated game assets or code is usually naive contamination from market place re-sellers that don't have a legal right to the kit-bashing content.
Very common to find pitch shifted audio getting resold on Unity store or web, images/3D with intact watermarks/meta-data, and scraped shader code. If you are credulous enough to buy files from a shady online seller, the copyright/trademark liability almost always falls on the person that publishes the IP. You can do everything above board, and still get sued as other license holders may register the asset signature with anti-piracy/YT bots.
For example, you could buy BBC/NBC/Sony libraries off ebay/web for a few bucks, but they will 100% still sue you into bankruptcy if you ever publish content with their media IP.
"AI" vector search can produce similar results of existing scraped works, and almost every LLM knows Disney Mickey Mouse. How LLM are trained means isomorphic plagiarism will occur regardless of data granularity. Just like how mp3 were a lousy highly compressed version of CD audio, the output does not have to be a verbatim copy to trip copyright and trademark IP laws.
Isomorphic plagiarism becomes less obfuscated when compiled into standard executable code. However, some firms actually used human session chat admins for data labeling/edits.
https://www.youtube.com/watch?v=o1iLPxwgV4E
People may still lose even if they are right. If you resell something you didn't actually make yourself or purchased royalty free perpetual rights from the authorized holder, than the liability still falls on you when content rights change.
This is why fun indie games sometimes disappear. Some may not like these rules, but most play by them to survive as a business. =3