Link to the challenge result announcement.
https://krita.org/en/posts/2025/monthly-update-27/?pk_kwd=KritaMonthlyUpdate-Edition27
Krita hosts these painting events monthly then gives the award to their best pick.
Here is the link to the winner, by Mythmaker.
https://krita.org/images/posts/2025/mu27_mouse_sage-mythmaker.jpeg
Anyway, out of curiosity, I ran it through https://wasitai.com/. This is what it tells me.
We are quite confident that this image, or significant part of it, was created by AI.
The problem with “AI detectors” is that they are, themselves, AI and highly prone to just “making shit up.” They cannot be trusted to be accurate and cannot even explain the “reasons” for their appraisals. These tools are hurting real creators, including artists and students, and should not be used IMO.
There’s a few things here that tell me it’s probably not copyright-theft-generated. The big one that’s easy to explain is the tail. The tail starts off from behind the mouse, snakes in front of the cloak and background (so far so good), but then, here’s the critical thing, passes behind the fern staff and continues on the other side of it, positioned properly and in continuity.
Copyright-theft-generators have tremendous problems with this because, as the chorus goes, they don’t understand anything. There is no mental model of “a tail” with them. There is no thought of a tail’s properties, so keeping a tail contiguous while passing across barriers is very hard for them.
In a previous version the author started the tail at a slightly different spot (hidden, behind the body ) too. He had time to refine it in the latest, featured version.
The artist has shared two iterations of their painting, and you can flick through their post history to see them honing their skills over five years, including more art on the same clothed-mice theme, also with multiple versions.
Using an AI tool uncritically on !fuck_ai@lemmy.world to accuse a genuine artist of using AI is not really in the spirit of the community, IMO.Edit: miscommunication – OP was spotlighting the mistake made by the detection tool, not accusing the artist of using AI.That’s what I said. I’m sorry if I made you think that I was “exposing a fake artist using AI.”
Ah, the way you worded your post definitely gave me the impression you were accusing the monthly contest winner of using AI, not criticizing the AI detection service! And from the comments, most others read it the same way.
Thank you for being kind.
I get that from other comments too. If you read carefully, you can see all I said is fact, i.e., what happened. I don’t use the place in OP to express my own opinion because it may form a bias.
Oh, and any one thinks I am stupid enough to “out” skilled, regular artists on one of the most popular drawing softwares’ forum?
I donate to Krita regularly by the way. I love Kiki.❤️
In this case, “fuck AI” fucks the tools, which is AI itself, try to detect and claim real people’s work as AI-generated.
And fuck this AI art trend as well.
I think you perhaps assumed that others have the same context as you, in terms of familiarity with the software, the forum/community, and the contests. Given that you knew Mythmaker is an actual artist, a couple of words in the affirmative would have made a big difference to the reading of the post!
WasItAI is a powerful AI-powered tool designed to detect AI-generated images. It helps users differentiate between real and fake images, ensuring the integrity of visual content. WasItAI uses advanced algorithms to analyze the characteristics and patterns within an image. By comparing these features against a vast database of real and AI-generated images, it can accurately detect the image’s origin.
There’ve been proofs (scientific papers) that what we call AI currently can’t really do that
AI detectors are not really accurate as I understand it.
Imagine the level of rage (or indifference) when the author sees the AI-detection tool put on judges on his hard work. One error in 100 is enough to drive one insane.
Especially if the image is poisoned.
OP is catching some grief for using an AI AI-detection tool, but I also use an AI AI-detection tool when I’m not sure, and I’m wondering we else we can do? The supposed AI image-generators are getting better, five fingers and all, and I’m not art critic enough to always be certain. It’s a problem that’ll only get worse as the fake artmakers get better.
I will say, AI AI-detection and then rejection is the only use I’ve found for AI.
The problem with AI detectors are that if you try to protect your art against being used in training, using nightshade for example, will make the detector say it’s AI enerated, even if it’s fullt hand crafted.
Posting about using an ai on this is community is certainly a choice, I’ll give you that.
I think they used nightshade or something, that makes it stand out as AI created to AI detectors
