A viral cartoon with someone else’s name in the corner
After the deaths of Dolly Parton and Tim Curry in late August, a single-panel cartoon showing the pair at the pearly gates spread across Twitter, Facebook, Instagram and Bluesky. One post of it drew 25,000 likes. In the bottom right corner sat the pen name BLOPER — the signature of Brendan Loper, who draws for The New Yorker.
Loper did not draw it and did not sign it. The image was generated by ChatGPT and first posted to Facebook by a Dolly Parton fan who wrote in her own comments that she had asked the chatbot for a New Yorker-style cartoon. Loper told Nieman Lab, which reported the story on 5 October, that his twin brother texted him about it first, and that strangers then emailed and messaged him asking whether it was his. He had been warned about a similar output in May.
More than fifteen cartoonists, and a logo nobody licensed
Nieman Lab’s Andrew Deck documented more than 15 New Yorker cartoonists whose signatures the image generator had attached to work they never made. Among those he found in online forums or produced in his own tests were Harry Bliss, Emily Flake, Joe Dator, Pat Byrnes, Peter Vey, Jason Adam Katzenstein, Liza Donnelly, Ellis Rosen, George Booth, Saul Steinberg and Warren Miller — the last two long dead, and Miller a contributor as far back as the 1960s.
Not every generated cartoon carried a signature, and some bore invented ones. But the real names recurred, and some outputs also carried The New Yorker’s logo.
Condé Nast, which owns the magazine, signed a multi-year content licensing deal with OpenAI in 2024. Its terms are not public. A New Yorker spokesperson told Nieman Lab that Condé Nast has never granted any language-model developer permission to train on its cartoons, and that reproducing the magazine’s logo is not permitted under any of its AI agreements either. The magazine’s cartoonists submit on spec, are paid only when a drawing is selected, and keep their copyright; the standard contract is not work for hire and does not cover AI training.
OpenAI calls it a bug
An OpenAI spokesperson told Nieman Lab the company’s focus is on tools that empower human creators, and thanked the community for flagging what it called bugs and unintended behaviour by its models. OpenAI declined to answer how the signatures entered its training data.
After Deck notified the company, ChatGPT began returning a warning on some New Yorker-style prompts about similarity to third-party content. As of publication it was still signing some generated cartoons with real names.
The cartoonists Deck spoke to drew a line between style and signature. Flake described the images as carrying the DNA of her work without looking like anything she drew, and compared the effect to being quoted saying something she never said. Loper put it more simply: “my name is my name.”
US copyright law has not settled any of this. A claim over style would be hard; a claim over a signature is closer to false attribution or impersonation, which is a different body of law and largely untested against image generators. What is already established is the market context: in July a federal court approved a $1.5bn settlement between Anthropic and authors whose pirated books were used in training, the largest in US copyright history. Katzenstein, one of the cartoonists whose signature ChatGPT reproduced, was part of that case.
