Nieman Lab, the journalism-technology analysis site affiliated with Harvard University, has reported that ChatGPT generated images that closely resemble the single-panel cartoons regularly published by The New Yorker. According to the investigation, the chatbot not only copied the magazine’s characteristic black-and-white, single-panel look with sparse backgrounds and ironic captions, but in several cases also appended signatures that match real New Yorker cartoonists.
Specific incidents and figures
Nieman Lab identified similar instances involving more than 15 New Yorker cartoonists, including Brendan Loper, Emily Flake, Harry Bliss and Joe Dator. One widely shared example highlighted by TechSpot showed a cartoon that went viral after the reported August death of Dolly Parton: the image depicts Parton entering heaven alongside British actor Tim Curry, who also died that month, and bears the signature “BLOPER.” That sign-off is used by Brendan Loper, but Loper told Nieman Lab he did not create the drawing.
Nieman Lab says this was not the only time Loper’s signature was copied; during the summer the cartoonist himself found multiple similar forgeries on a Reddit forum focused on ChatGPT.
Why does this happen?
The likely cause is how image-generating models are trained: they learn patterns from massive collections of images scraped from the internet. If a cartoonist’s body of work frequently pairs a consistent drawing style with a recurring signature, the model may infer that the two elements belong together visually and reproduce both.
Andrew Deck of Nieman Lab notes that signatures today carry more than copyright information: in an era when illustrations and cartoons are often removed from their original context and reposted without permission, signatures can act as de facto watermarks that tie an illustration to its creator.
Responses and responsibility
A New Yorker spokesperson told Nieman Lab that the magazine never licensed its cartoons for training large language or image models. OpenAI declined to detail how its models might learn or reproduce signatures; it characterizes such outputs as "unwanted behavior" or bugs in the model.
After Nieman Lab alerted OpenAI, ChatGPT began showing a warning for certain prompts requesting New Yorker–style cartoons: “This prompt may violate our rules about resemblance to third-party content.” Nevertheless, Nieman Lab reports that some generated images continue to carry names or signatures resembling those of real New Yorker cartoonists.
Why it matters
The issue raises legal, ethical and trust questions. Creators and publications worry that their styles and names can be improperly associated with AI-generated content, and audiences may struggle to distinguish human-made from machine-made works. Nieman Lab’s findings suggest that the current behavior of image-generation models—producing forged signatures and imitating distinct artistic styles—may require further technical safeguards and regulatory attention.
