[Pfeil]

First of I think it is nonsense as I never got any NSFW content with filters enabled, and I actually always disable them and still get no NSFW content if I do not search for topics that can at least imply it.

But the main point here is that this is something generated in a chat-like product. The task and expectations and level of control are completely different. I am not saying it is impossible something like this comes up in a model even with good training data, but even in this unlikely case, this would be easy to catch with a filter.

AI is not search. It is a likelyhood-based text generator which can be useful for finding good terms for search and a lot of other uses, and while people use it for search, the mode and UX is completely different. That is why people use it as such!

If a Restaurant cannot provide proper hygiene, it will be forced to improve or be closed if noticed. And for Grok (and others) hygiene is in question due to such incidents. Society needs to discuss metrics for AI chat hygiene so we can have consistent law enforcement to shut down services like these when they are not hygienic enough.