Your first comment was incredibly vague… I was responding to this part:
Glad corporations get the power to make these decisions.
However, a high false positive rate is different than assuming every post is “guilty until proven innocent”, and they aren’t mutually exclusive either. Current example here would be the automated removal of CSAM on Lemmy. A model was built to remove CSAM and it has a high rate of false positives. Does this mean that it assumes everything is CSAM until it’s able to confirm it isn’t? No. It could work that way, that’s an implementation detail that I don’t know the specifics of, but it doesn’t necessarily mean it does.
But really, who cares? The false positive rate matters for site usability for sure, but the rest is an implementation detail in an AI model, it isn’t the court of law. Nobody’s putting you in Facebook prison because they accidentally mistook your post for rule breaking.