This is a bit of a re-report of this bug due to seeing it happening again:
So yesterday, I was playing Roblox when I found someone bypassing a swear acronym into chat, even though this whole thing should have been fixed. So after that experience I tried to use these swear acronyms in my dev testing game with no one around (for bug reporting purposes) and alarmingly, a lot of these acronyms weren’t filtered as shown here:
if you are wondering what these acronyms mean I listed them in this dropdown menu:
Acronym meanings
stfu, this means “Shut the ‘f-word’ up”.
syfm, this means “Shut your ‘f-word’ mouth”.
sybau, this is the one that was partially fixed but not quite as it can still be bypassed by placing emojis next to it without a space, this means “Shut your ‘b-word’‘a-word’ up”.
fafo, this one hasn’t seemingly been used that much but isn’t censored and I included this for extra measures, this means “‘f-word’ around and find out”.
fml, also isnt used much but I again included it for extra measures, it means “‘f-word’ my life”.
tf, not much to explain here, it is basically a shortened version of the already banned acronym “wtf” which means “what the ‘f-word’”.
As we can see, these acronyms are very inappropriate for Roblox and shouldn’t be allowed due to the language used.
Expected behavior
All of these acronyms should be censored and the emojis should not be able help bypass already banned ones like “sybau” into the chat.
I’ve noticed that a lot of acronyms with swear words are not being censored despite a lot of them mainly being used in disrespectful ways. I’m not sure if this is a part of Roblox’s chat filter loosening, but I don’t think they should be able to say this stuff…
Sorry but, what’s the point of censoring the words if you’re just typing “f-word” instead of actually typing the real word?
Anyways, my guess is that, because Roblox filters are AI-based, they need to constantly train and fine-tune it to account for new words.
But as the filter model is fine-tuned, it may adjust some old weights and biases that could possibly result in older words unintentionally being unfiltered again.
I don’t want to get flagged abusively because I don’t trust them → . anyways I do wonder if those words could just get put on a sort of “permanent no use blacklist” if that exists in the filtering system