Literature.cafe's meta community for announcements, requests, bug reports. etc etc.
Please follow the instance's rules
Hi, the admin of feddit.nl here.
Made an account on lemmy.world to make sure this message gets federated.
I've changed the registration to approval to avoid some spam.
I've also found a second admin who will support me in cleaning up spam and csam stuff.
I'm open to more admins, but it can be challenging to find trustworthy people on the internet these days.
@[email protected] does this mitigate your worries a bit?
Copying here what I put in matrix: I understand, I will refederate and remove the censure as well. Sorry if this put a bit of stress on you. Just knowing you were making a good faith attempt to find an admin of some kind would have been enough tbh, it's a lot of trust and lemmy needs more granular permissions for site staff. Having it be admin or nothing isn't really helpful at all.
Man it sucks a fat dong that you guys have to deal with this shit. I appreciate your efforts. Keep it up boys, you’re killing it.
I’ve begun to love Tetris again so that’s something from this I guess
I'm sadly to the point where I will automatically defederate any instance that has open signups and am seriously considering moving to the allow-list model. feddit [dot] nl was still on my blocklist from the last time their open signups caused problems, so I was spared this round.
just to be clear i dont blame their site admin for this here but i really would feel much safer with an instance of their size having at least one other admin if they are able to or at the very least making it clear in communication they are reporting these posts to the authorities
will let y'all know what comes of that 🫡
pictrs-safety is ready to host. Just need someone to dockerize to make it easy for y'all to catch CSAM before it's even federated.
The fact it requires a GPU isn't really feasible to host for me, as well as the fact idk how comfortable I am with potential legal liability it might open.
I am planning to add a feature to scan for porn (not csam) via ai horde which will bypass the need for gpu if you want to use that m
Great work!
How does this work? I get that it's a docker setup, but I mean what does this actually do and how, for picture safety against CSAM?
It scans each picture with AI before it's uploaded to determine if it's potential CSAM, and if it passes the thesholds, blocks the upload
What is the AI engine? Who controls the models for it? Does it interface with any known CSAM databases?
You can read about it here https://github.com/db0/fedi-safety
That links to an entirely different repository/software. Does pictrs-safety require fedi-safety? Are they independent ? Does pictrs-safety upload to fedi-safety on... your servers? Someone elses running fedi-safety? Does the user using pictrs-safety require a fedi-safety installation working too?
Lots of unanswered questions.
They're both independent from each other but meant to work together. However you can optionally plug anything else you want on pictrs safety.
You should run on your own systems.