I'm not an expert, but aren't pretty much all neural networks that are used for classifying images prone to adversarial attacks and general weird conclusions? I

I'm not an expert, but aren't pretty much all neural networks that are used for classifying images prone to adversarial attacks and general weird conclusions? In this case it comes up with something offensive, but I'm sure there are plenty of other cases where it's just incorrect etc. I'd imagine the only solution would be to have a human editor before these are posted, or just not use this technology in that kind of context.