It wont do like "hey screw you an AI decided that your comment is bad" rather only if the comment swears hard like N-Word, Hard Insults etc. he steps in hard and do something.
Automod is one of the best tools for such things because it is not biased of anything it just has that insult list and filters it out.
Cougar: I'm just indicating who's talking; I share the account with the twelve other people rattling around in this noggin, so disambiguating makes sense.
Pearl: And our concern isn't someone spamming the bot, our concern is someone outsmarting it. For example, figuring out the list of "tripwires" for a given server and baiting otherwise innocent posters into getting bonked.
It would probably not be 1 Bot for 99999 Comms rather 1 for 5 communities ( pre selected ) and those are hosted by someone who is in their communities mod. And they can just edit the bot to add or remove bad words or filter out other factors. And how do you share an account AND a comment? Or do you share it on 1 pc too?
Pearl: We literally live in the same brain as each other. We physically cannot separate, so we figured it made sense to just treat all accounts as collective.
If a malicious actor wants to spam this. I am sorry to tell you he doesnt need a bot. Lemmy has no real anti bot measures ( other than ratelimit ) So just a simple bot net or just proxies and spam the heck out of a lemmy instance. The lemmy instance is going to have some spam.