2026-09-26 04:32 UTC
@juergen_hubert@mementomori.social I'm pretty sure that the safety tools will require a great deal of rethinking.
Currently safety is dumped as a workload onto instance owners; instance owners who are not given any reasonable tools for doing that job. I'm an admin over here, for example, on gamerplus and in July to August we got flooded by a bunch of AI-based bot-made accounts. It was a daily chore to go through the new accounts and decided quickly if a new account was a fake bot account or a real one.
🧵 ▶️
Replies (1)
-
@zdl@gamerplus.org 2026-09-26 04:36
@juergen_hubert@mementomori.social Of use would be databases of known bad ID patterns, known bad email hosts, known bad actors, etc. that we could use to just automatically nix accounts without even having a human in the loop except for checking out the final decision. We're small enough that harassment isn't much of an issue; we had one person once make an account as part of a process of harassing people in a personal crusade. But I can see how a larger instance with more breadth could get hammered by it. 🧵 ▶️