@CursedSilicon It kind of depends on what _they_ mean by “AI upscaled”, as there are many approaches. But I think if you have ethical issues with sourcing training data, probably any of these are disqualifying.

Personally, I’d avoid it for the same reason I distrust most genAI content - because I don’t believe the operators have audited the output for more than simple, cursory cases, which means it’s your job to tolerate the errors.