This is a Lemmy Community for perchance.org, a platform for sharing and creating random text generators.
Feel free to ask for help, share your generators, and start friendly discussions at your leisure :)
This community is mainly for discussions between those who are building generators. For discussions about using generators, especially the popular AI ones, the community-led Casual Perchance forum is likely a more appropriate venue.
See this post for the Complete Guide to Posting Here on the Community!
1. Please follow the Lemmy.World instance rules.
2. Be kind and friendly.
3. Be thankful to those who try to help you.
4. Only post about stuff related to perchance.
5. Refrain from requesting Prompts for the AI Tools.
text-to-image-plugin and ai-text-plugin) e.g. "What is the good prompt for X?", "How to achieve X with Y generator?"6. Search through the Community Before Posting.
I'm so angry I'm shaking while typing this. Many others and I have reported this one alone, and it's still here! It looks like it's gotten more tabs to try to hide.
Please. Ban this filth. This is NOT okay.
There is no valid reason for children to be made on a generator that allows NSFW. And some of the tabs are filthy when I think of them in the context of CP.
All these pedophilic plugins are just an add-on to form a request to the AI, which already has mechanisms for drawing child pornography. Remove the add-on — the same request will be typed manually, and that's it. The AI is just a tool. You're now trying to blame the paint seller because an artist drew smut. To achieve what you want, you shouldn't remove plugins — you need to add a core AI query like 'analyze the request and respond if it doesn't result in obscenity.' But that requires allocating additional resources for processing this request. Judging by how systematically the AI's performance declines here, the admins are too stingy with those resources. Well, here we are.
You have to understand that there's not a lot of people who work on the Perchance website so it's going to take time for stuff to get taken down. I'd heavily suggest you get in contact with an admin or someone who works with Perchance.
Okay but why is it there in the first place? The whole world is theatrically crying "The Children!" as they gate the internet behind identity. So why was this even made during this "sensative time"?
Do you think this was directly created by the admins? This is how it goes when you open a service to the public, unfortunately. All any platform can do is try to police retroactively. They can try to put in some safeguards, but that's an eternal game of cat and mouse because slimy fucks will study how to circumvent it, admins invent new safeguards, repeat loop.
Idk I think you can screen for that on your hosted service pretty easy. It has to store the data somewhere and it can be checked. Dont act like its impossible its not it is probably costly and that's why this is possible.
Yeah, they can scan, and as soon as they start doing that, people who want to host it there anyway will start testing things to see what triggers removal and use what they learn to circumvent the scan. The host then refines the scan and catches the new tricks. Then belligerent actors will start probing for even newer tricks, which will eventually be put in new scans, which will eventually be circumvented by the newest tricks.
That said, maybe Perchance could reasonably be doing more. I don't know from this one instance. But this sort of thing is very much not a solved problem, and it's not so simple as "just scan for it", so I'm hesitant to blame Perchance without evidence it's more systemic.
It shouldn't be possible to make CSAM on Perchance, even if generator authors have combinations of dropdowns which combine age keywords with NSFW keywords. If there are combinations of keywords that produce CSAM, then that's a bug with the text-to-image-plugin that should be reported (you can DM me here). Reporting generators that are clearly designed to generate CSAM is helpful, but less so - it's more whack-a-mole, rather than solving the underlying issue.
If your concern is that people are generating non-CSAM pictures of children, and are possibly being sexually gratified by them (e.g. child beauty pageant pictures or similar), then it's quite hard to prevent that at the text-to-image-plugin level, since it's about the mind of the observer rather than the image itself. There's a surprisingly long tail of things in this category that even extend to semi-nude children in non-sexual situations - e.g. cherubs / renaissance paintings, tribal settings, swimsuits, and even semi-nude children with semi-nude adults in the same picture - e.g. breastfeeding, childbirth.
That said, there is certainly still a lot of room for improvement on the current CSAM-prevention system. Taking it to the next level is unfortunately more like a research project (mainly becasue of how cheap I need to run it) than simply flipping a "enable stronger CSAM filtering" switch, so I don't have a clear ETA at this point, but in the meantime it's helpful if people can report loopholes in the current system.
Is this a public generator or a custom one? If you check the source in the edit, there's a regex for catching stuff like that generally but if its a custom generator then you should report the generator and not the platform.