Several months ago Beehaw received a report about CSAM (i.e. Child Sexual Abuse Material). As an admin, I had to investigate this in order to verify and take the next steps. This was the first time in my life that I had ever seen images such as these. Not to go into great detail, but the images were of a very young child performing sexual acts with an adult.
The explicit nature of these images, the gut-wrenching shock and horror, the disgust and helplessness were very overwhelming to me. Those images are burnt into my mind and I would love to get rid of them but I don’t know how or if it is possible. Maybe time will take them out of my mind.
In my strong opinion, Beehaw must seek a platform where NO ONE will ever have to see these types of images. A software platform that makes it nearly impossible for Beehaw to host, in any way, CSAM.
If the other admins want to give their opinions about this, then I am all ears.
I, simply, cannot move forward with the Beehaw project unless this is one of our top priorities when choosing where we are going to go.
First of all, I’m so sorry that you have been exposed to such horrors. I hope you can handle that, or find help to.
I don’t have a solution, I’d just like to share some thoughts.
Some people suggested that AIs could detect this kind of content. I would be reluctant to use such tools, because lots of AI projects exploit unprotected workers in poor countries for data labeling.
An zero-image policy could be an effective solution, but it would badly impact @dyi@beehaw.org, @creative@beewah.org and @greenspace@beehaw.org.
correct me if I’m wrong, but on the fediverse, when a picture is posted on an instance, it is duplicated on all federated instances? If I’m right, it means that even if beehaw found a way to totally avoid CSAM posting, you could still end up with duplicated CSAM on your server? (with consequences on your mental health, and possibly legal risks for owning such pictures)
Kind of. It duplicates on all instances that subscribe to the community where it was posted to. Behind the scenes, Lemmy makes each community a “user” that boosts everything posted to that community. That content, is only getting pushed to instances where at least one user has subscribed to that community/“user”, then any included images get cached. So if nobody subscribes to a federated instance’a community, none of the content gets duplicated.
The biggest problem right now are users with “burner accounts” who exploit instances with free-for-all registrations, to push content to communities that have subscribers from as many different instances as possible… possibly “lurker” accounts created by the same attacker just to subscribe to the remote community they’re attacking and have the content show in the default “All” feed of all instances.
There are some possible countermeasures for that:
Thank you so much for all these explanations! I didn’t know the communities/users were so important in the system.
I thought that a duplicate of each post on a instance was automatically sent to all federated instances, and I wondered how the servers didn’t get overloaded by the global activity.