Using this approach, I've now turned on automatic scanning of new uploads.
What this means for you is that occasionally you will upload an image for a post and it will stop working after a bit. C'est la vie. Just upload something else. Changing format or slightly altering the image won't help you.
Also, sometimes you might see missing thumbnails on post from other communities. Those were the cached thumbnails hosted by us. The original images should still work in those cases.
Unfortunately this sort of AI scanning is not perfect and due to the nature of the beast, it will catch more false positives but to an acceptable degree. But I find that this is OK for a small social network site run as a hobby project.
They know these servers are being run by amateurs who lack the capacity to deal with it at any real scale or for long term.
So they are being targeted.
That being said there have been a large vocal minority who get shit on everytime they point out that lemmy and it's top servers are 100% not prepared to protect privacy nor have the ability to respond to this shit. This was predicted but the community is absurdly against paying anybody to run a proper fucking server employing professionals and giving a fuck about how to properly deal with csam. "I'll donate rather than subscribe." Sure you will Chad, sure you will. At best you'll donate they'll $20 total which will cover maybe a month of compute and bandwidth.
This place will self destruct in a few months when volunteers are sick of looking through fucked up images for no pay or even professional mental support which is like day fucking 0 shit for anyone dealing with scam. Even fucking Reddit has a responsive team to come help out mods with this shit.
You have a common-sense point there, and so what should we (if not all Lemmy and Kbin online rafts and their denizens, most of them) do at this point?
Have at least one competent, small-enough mental health team per instance or three, whose primary job functions alike that of an active mental health hotline, albeit geared towards the Fediverse citizen in need, admins, moderators, community builders, coders, creatives, and lurkers alike, but make sure they don't overstep their bounds a bit too much, and that they also connect with others representing other instances, keeping in mind the right to privacy and all other related human rights to make sure the ecosystem stays healthy at all facets, not just in the online, but also in the "outside world" sense.
For that, diversity, genuine compassion and science-and-humanity-based objectivity are also needed for those willing to take up those roles.
Catch and hunt these degenerates. I'd hope law enforcement would get involved. If these sick fucks have these images and have posted them hopefully they can track them and imprison them
I happen to agree with this somewhat blunt individual here. They're annoying, and though I'm not generally in favor of censoring even people as abhorrent as they, when it becomes so prevalent and aggressive that it is tantamount to spam, I think it is warranted. The genocide denial and all that is just icing on the cake to me, frankly.
This is dang impressive, nice! As someone entirely unappreciative and unaware of Lemmy's backend, is it likely a tool that other instances could/would make use of? Or is it too faffy to redeploy for other instances?
This is perfect moment / place for interpol and other organizations to catch some pedo mother fuckers . Those people are absolute human trash. Also Expect heavy vigilance from big eyes over the fediverse.
Very cool! Personally I'd think even blocking any pictures of children on the website, as those are almost always posted by adults without real consent and rest of the time children who do not understand the consequences of uploading their photos to internet.
and due to the nature of the beast, it will catch more false positives but to an acceptable degree
What is this "acceptable degree"? Where is it documented?
What is the recourse for the uploader in case of a false positive? And no I don't mean "upload something else", I mean what do you answer to "my legit content is being classified by a shared internet tool as CSAM, of all things".