Reading
ChatGPT Is Signing Real Cartoonists' Names. That's Not 'Style'
ChatGPT put real New Yorker cartoonists' names on AI cartoons they never drew. Why that's worse than style-copying, and what to check before posting.
Prices last read 2026-10-20
Short answer
Yes. In some images, ChatGPT has been writing real cartoonists' signatures onto cartoons they never drew.Andrew Deck at Nieman Lab reported on October 5, 2026 that OpenAI's image generator used the signatures of more than 15 New Yorker cartoonists, with no permission and no payment. Names include Brendan Loper, Harry Bliss, Emily Flake, Joe Dator, Pat Byrnes, Peter Vey and Jason Adam Katzenstein. After Nieman Lab contacted OpenAI, ChatGPT began warning on prompts for New Yorker-style cartoons. When the story was published, though, some generic cartoons were still coming out signed with real artists' names.Source: Nieman Lab (Andrew Deck): ChatGPT is adding real cartoonists' signatures to fake New Yorker cartoons, full text read via Nieman Lab's RSS feed · 2026-10-20
How a cartoon signed 'BLOPER' went viral
According to Nieman Lab, the cartoon that set this off featured Dolly Parton and Tim Curry and was signed "BLOPER," the signature Brendan Loper uses on his New Yorker work. Loper didn't draw it. ChatGPT made it after a Parton fan asked for a New Yorker-style cartoon, and the fan posted it to Facebook. From there it spread to Twitter, where one post got about 25,000 likes. People started asking Loper whether it was his, and he called the experience "very surreal." Deck then ran his own tests and got further signatures, including George Booth, Liza Donnelly, Ellis Rosen and Saul Steinberg.
Source: Nieman Lab (Andrew Deck), read via Nieman Lab's RSS feed · 2026-10-20
OpenAI's statement to Nieman Lab said it values the community flagging bugs and unintended model behavior. OpenAI wouldn't say how its models might have picked up signatures from The New Yorker's archive. Its fix so far is a warning on New Yorker-style prompts: "This prompt may violate our guardrails concerning similarity to third-party content."
Source: Nieman Lab (Andrew Deck), read via Nieman Lab's RSS feed · 2026-10-20
Our take: copying a style is debatable. A fake signature isn't.
People can argue all day about whether AI copying a house style is fair use, homage or theft, and they will. This isn't that argument. A signature isn't part of a style. It's a claim about who made the work. Loper described his as a "certificate of authenticity," and that's exactly it. If a tool writes a real person's name on a picture they never drew, it is making a false statement about authorship. Katzenstein's description was "slop that's pretending to be me." Calling that a bug undersells it.
The fix is weak, too. A warning that only fires when someone types "New Yorker" puts the burden on the user's wording, when the problem is what the model writes in the corner of the image. Nieman Lab found real names still appearing on generic cartoons, so the guardrail is covering the prompt and missing the output. If you pay for ChatGPT, you're the one posting these images under your own name. You should expect OpenAI to filter real signatures out of the image itself, not just flag a few keywords.
Represent that Output was human-generated when it was not.
OpenAI Terms of Use, in its list of things users may not do (effective January 1, 2026)Source: OpenAI Terms of Use (effective January 1, 2026), read via a plain-text rendering of the official page · 2026-10-20
Why this matters to you: OpenAI's terms forbid users from presenting Output as human-made when it wasn't. A cartoon with a real cartoonist's signature on it does exactly that, even if you never asked for the signature. The same terms also say output "may not be unique" and other users can get similar results, so the same fake signature can show up in more than one person's images.
Source: OpenAI Terms of Use (effective January 1, 2026), read via a plain-text rendering of the official page · 2026-10-20
Before you share an AI cartoon
- Zoom in on the corners and margins. Look for anything that reads like a signature, initials or a pen name.
- If you find a name you didn't ask for, don't post the image. Regenerate it, or remove the signature before it goes anywhere.
- Say clearly that it's AI-generated. Don't let a fake signature imply a human drew it.
- Avoid prompting for a specific magazine or named artist if you plan to publish the result.
- If you get a real artist's name on an image, report it to OpenAI as unintended behavior. OpenAI told Nieman Lab it appreciates that kind of flag.
Who's actually exposed
Casual users posting a joke are the ones most likely to get caught out, because a viral cartoon gets credited to the real artist almost immediately, like BLOPER was. The artists take the reputational hit, and the person who posted it looks like they faked it.
Optional shopping: see a seller listing for ChatGPT plans. It's a third-party listing, not official OpenAI access, and not our endorsement.. That link carries our partner code Sk52g, which is how a purchase is attributed to this site. Promotion code UPD3S.
Sources (read 2026-10-20)