I created a badge icon for one of my checkpoint obelisk/statues using a screenshot of the model itself, the image got reviewed by the AI and got removed for Discriminatory Content and banned me for a day.
Now, normally I didn’t expect this to be an issue as from what I can tell, this is a clear false positive from the AI moderation, and a simple appeal would clear it up.
I am astounded as to how the moderation managed to look over it in less than a minute and deny it. Is the appealing of reviews also AI moderated or do we have real humans manually reviewing it? In my opinion, any human would look at this image and not see anything wrong, especially ‘Discriminatory content’ Perhaps I am missing something. I would appreciate some clarification as to how this gets flagged in order to avoid this in future. Thanks
The person doesn’t consider your appeals, it’s done by the same crazy AI Roblox that deleted your icon. Regarding the icon, when I uploaded the maze, my image was also marked in a similar way by AI. I see some similarities at the bottom.
From an isometric perspective, the two purple thunders on the obelisk look very similar to the logo / insignia of the Schutzstaffel organization from Nazi Germany, could possibly be why the AI flagged it?
This might be it, would make sense as to why the maze from zombi_vadim got flagged aswell; AI recognising the wrong patterns. Next time, I’ll try reuploading without them.
I ran out of appeals, they denied all of them lol. It would be helpful if it told me what was getting flagged so I knew what to change rather than taking a random gamble and hoping I don’t get banned
Welcome to Roblox. Here you can get banned or deleted for no reason for anything, anytime.
Many download resources from alternative accounts in an attempt to avoid AI moderation on the main one.
Seems like AI model overfit, maybe on geometric patterns, should bug report it so people have somewhere to be pissed about shitty under-tested AI moderation. These things should have continuous rolling false positive detection tests and a better appeal pathway.
