Artist: kovar (locked)

Staff members involved: Lafcadio, Xch3l, Bearnaise, Song, Versperus, Recursion, Wandering Spaniel, Rainbow Dash, binaryfloof, NotMeNotYou, abadbird, spe

At this time, the Bladerunners' stuctured investigations did not exist: the practice of collecting samples from AI-only art sites and interviewing appellants to determine certain facts about their workflow did not exist. Kovar's case was first handled in response to an increase in commenter activity around the artist, accusers and defenders at each other's throats. Specific complaints by staff at the time were:

• "there's some inconsistencies around the teeth where some blend into skin"
• "breath clouds look like overlaid stickers"
• "shading between breasts looks smudged, in contrast to the cell shading of the rest"
• "similarly, dark strokes seem like actually drawn over as there's lighter edges around some"
• "maybe it's me but texturing is also inconsistent, if they took time to add texture to the thighs and torso why not do the arms too?"
• "[jpeg shenanigans]"
• "several smudges that not even filters could do around the face"
• "a commenter also pointed out the teeth and i think it's pretty damning [crop of post #4665068 highlighting several regions where there is no lineart separation between the teeth and the mouth.]

The discussion was immediately derailed by Discord introducing a poll feature. Two days later, Mairo posted in a janitor channel, bringing attention to two of Kovar's posts with a comment showing interest in removing the flags if no further investigation was made. Strikerman brought up Bearnaise as somebody investigating the posts, and a back-and-forth between Mairo and Bearnaise ended in the posts' deletion.

Starting 2024-05-29, Versperus viewed a ticket including a link to a detailed Google document, titled "KOVAR SHARK IS FAKE". A poll was opened to try and quickly learn janitors' assessment of Kovar, with two respondents: both Song and Versperus voted "I don't know". Over time, more attempts were made to coordinate staff responses to the accusations, handling misflags of new posts as "Previously deleted", revisiting the discussion as more and more staff became aware of the case, etc.

2024-09-12, the artist sent a DMail appeal to meowmcmeow inquiring about these "Previously deleted" flags, correctly stating that it was impossible unless somebody else had uploaded their art in advance. NotMeNotYou asked if anybody had performed more detailed investigation after the first set of deletions. Strikerman highlighted a sample from the artist, and NotMeNotYou gave the go-ahead to purge their gallery. abadbird highlighted more samples of suspicious quirks in Kovar's output.

With the matter seemingly resolved, binaryfloof tried to draw attention to another hanging AI accusation. This unrelated artist was later subject to a conviction based on an AI coloring disclosure.

2024-10-04, the matter was raised again by abadbird after somebody had again flagged many of Kovar's new posts. Lafcadio responded that Kovar's posts seemed to satisfy the bare minimum requirements for AI paintovers (at the time, full paintovers were being allowed.) This specific policy was discussed with comparisons to one artist who did make full paintovers of AI images, and another who did not. This discussion wrapped around to post #3747581 as the golden standard for an AI-assisted work, being based on an AI reference but not incorporating any AI coloring or lineart, the idea that certain full paintovers still might not have artistic value, and trying to tackle how to get different janitors to enforce a similar set of standards.

Later on, as e621's handling of AI posts changed, and as AI paintovers were disallowed entirely, there was no further activity related to Kovar for a while. 2025-06-07, Bearnaise received a DMail after she had approved a post by Kovar. Lacking the time and energy to constantly investigate Kovar's work, Lafcadio was notified directly in order to continue investigating, as by this point his expertise had already been established by a staff thread in April 2025 where he detailed various AI models, their tendencies, and specific sites where AI-generated art could be studied. Later in the same day, the user who had initially sent a DMail to Bearnaise rescinded their accusation based on Kovar proving the existence of layers in their work, but this only proved that Kovar's workflow was more sophisticated than posting unmodified AI generations.

e621's approach to AI investigations had fundamentally changed, so there was cause for further investigation. He noted one of his usual heuristics that could not be applied due to the nature of Kovar's portfolio, a few others that yielded more suspicious results, and comparisons against similar artists already being formally accused of AI usage.

2025-09-09, Mairo posted: "So like, what if I threw all of you a curveball and said, that Kovar is legit?" Mairo shared anecdotes about speaking with the artist, binaryfloof discussed needing a detailed case file to review the facts, and it was at this point discovered that the Google document previously used against Kovar was deleted. The artist opened a #private-help ticket on 2025-10-18 for a formal appeal, providing a sketch, two programmatically-generated timelapses (this in particular allows an image to be traced without making it visible in the timelapse), and giving testimony.

Lafcadio restarted the investigation from scratch, pulling up tweets by certain Twitter users and mirroring much of the evidence. He began noting certain bits of unusual metadata, unusual style shifts, and a particular image highly consistent with known AI generations. Additionally, a few posts were highlighted where certain details made it obvious that Kovar was redrawing these on top of finished AI images, and again brought up the prior complaints of teeth merging or not merging with the mouth at seemingly random. There was also testimony by a commissioner who noted Kovar's failure to provide sketches, and a transcript was made of the deleted Google doc by pulling up a text-only copy from archive.org.

Even by modern standards, Kovar's guilt was still extremely likely. spe and Rainbow Dash concurred. Lafcadio gave a final response on 2025-11-01, explaining that several posts had completely unmistakable signs of AI coloring, unable to be disproven by the submission of sketches and flats because they did not actually demonstrate the suspicious coloring elements. He also noted that other sources repeatedly complained about an inability to get sketches from Kovar, suggesting that for much of their portfolio, these sketches simply did not exist. This reply ended with: "We've got about 130 finished posts from you. Would it be possible for us to get 130 sketches from you, or at least something reasonably close to that number?"

After 8 days of complete silence from Kovar, the ticket was closed. No attempts to reappeal have been made since then.