ChatGPT Put Real Artists’ Names on Fake Cartoons

in August, a ChatGPT-generated cartoon went viral, depicting Dolly Parton and Tim Curry arriving at Heaven’s gates. The prompt was simple: a “The New Yorker-style cartoon.” But this wasn't just a generic imitation; the image bore a signature in its corner, unmistakably resembling that of cartoonist Brendan Loper Brendan Loper Loper.

Fans, and even Brendan Loper Brendan Loper Loper’s twin brother, quickly reached out, asking if he had drawn the piece. The answer was a definitive no. This was not Brendan Loper Brendan Loper Loper’s work, yet his identifiable mark, “B. Brendan Loper Loper,” was appended to it.

The distinction here is critical. Generative AI has long mimicked artistic styles, but this instance crossed a line. ChatGPT was not merely producing a drawing in the broad style of The The New Yorker; it was attaching a specific, working artist’s name to work he did not create.

This wasn't an isolated incident. An investigation by Nieman Lab later confirmed that ChatGPT had used the authentic signatures of more than 15 The New Yorker cartoonists, including deceased legends like Saul Steinberg. The models, once producing illegible scribbles, had evolved to render legible autographs, directly from style prompts. This isn't just style mimicry; it’s digital identity theft.

This wasn't one artist—or one harmless glitch.

That viral image was no isolated incident. Nieman Lab reported that ChatGPT-generated cartoons bore the signatures of more than 15 The New Yorker cartoonists, including Emily Flake, Joe Dator, and even Saul Steinberg, who died in 1999. Older image models often produced garbled, illegible marks, but these new outputs rendered clear, identifiable names, directly associating fake cartoons with real artists.

Attribution matters, and these artists were not amused. Emily Flake likened the experience to having a quote falsely attributed to her—a direct assault on her creative identity. Joe Dator, a The New Yorker contributor for over 20 years, put it more pointedly: "I've had people hack my credit card, and that feels less like a violation than this, because when they hacked my credit card, they didn't dress up like me."

The stakes are clear. This isn't just about style mimicry; it's about digital impersonation and the theft of professional identity. When an AI can convincingly forge a signature and attribute its output to a living or deceased artist, the line between inspiration and infringement dissolves, leaving a trail of confusion and outrage.

The plot thickened when OpenAI and Condé Nast, The The New Yorker's parent company, announced a content licensing deal. This agreement, however, explicitly did not greenlight training on The The New Yorker cartoons or permit using its artwork for image generation. That’s a crucial distinction, and one OpenAI has yet to clarify.

Cartoonists typically retain individual copyright to their work, a standard industry practice that makes the AI’s signature mimicry even more egregious. OpenAI has remained silent on how its models learned to reproduce these specific signatures, leaving a gaping hole in the transparency of its training data.

This isn’t just about style; it’s about attribution and intellectual property. The output raises serious questions about the provenance of OpenAI’s training data, even if reporting doesn’t pinpoint the exact source for any particular signature. For more on this digital identity theft, see ChatGPT is adding real cartoonists' signatures to fake The New Yorker cartoons - Nieman Lab.

The situation highlights a critical disconnect: a broad licensing agreement doesn't automatically grant carte blanche to exploit individual creators' unique identifiers. This isn’t a harmless glitch; it’s a direct challenge to the very concept of artistic ownership and personal brand.

OpenAI’s reported response to the signature scandal? A prompt for a “The New Yorker-style cartoon” now triggers warnings about similarity to third-party content. While a nominal guardrail is better than none, journalists found this behavior inconsistent—hardly a robust defense against reputational damage.

The stakes here extend far beyond style imitation, which itself often skirts copyright. A recognizable signature creates confusion about authorship, endorsement, and identity. This raises distinct legal questions, including potential violations of right-of-publicity and false endorsement, that a simple content warning cannot resolve.

Image-origin metadata may show a picture originated from OpenAI, but it cannot show that the named artist approved it. When fake cartoons bearing real artists' names spread, accountability demands more than a digital watermark. What specific remedies and safeguards will protect creators when their identities are co-opted and their reputations put on the line?

Did Brendan Loper draw the viral ChatGPT cartoon?

No. The cartoon was generated by ChatGPT, but it carried a signature resembling Loper’s.

How many New Yorker cartoonists’ signatures were reported?

Nieman Lab documented more than 15 cartoonists whose names appeared in generated cartoons.

Did OpenAI’s deal with Condé Nast authorize training on cartoons?

The New Yorker said the deal did not authorize training on its cartoons or using its artwork for image generation.

Can an AI-generated signature prove who made an image?

No. A visible signature can be fabricated, and provenance metadata does not establish that the named artist authorized it.