More than 15 New Yorker cartoonists, including Harry Bliss, Emily Flake, Joe Dator, and Jason Adam Katzenstein, have had their signatures or pen names appear on AI-generated images they never made, according to a Nieman Lab investigation into ChatGPT’s image generation system.
Cartoonist Brendan Loper’s pen name, “BLOPER,” appeared on a ChatGPT-generated cartoon depicting Dolly Parton entering heaven alongside actor Tim Curry. Loper said he did not create the image, yet viewers contacted him directly, believing it could be his work. That cartoon spread widely on social media before the confusion became public.
A Signature Is Not a Style Detail
False attribution creates a harm distinct from visual style imitation, turning an AI output into content that appears to claim human authorship.
Visual style imitation raises its own legal and ethical questions, but a forged-looking signature creates a separate problem. Signatures function as provenance signals: audiences read them as confirmation that a named human artist made the image, not as decorative marks that happen to resemble a real person’s handwriting.
For the cartoonists named in the investigation, that confusion could create reputational risk. Work they did not create now circulates under their names, potentially shaping how audiences, editors, and future clients perceive them.
A Licensing Deal That Did Not Cover This
Nieman Lab’s own prompting tests found that ChatGPT could readily produce New Yorker-style cartoons on request, and a platform warning did not stop false attribution.
After Nieman Lab contacted OpenAI, the system reportedly began displaying a warning about similarity to third-party content. AI-generated images continued to carry real cartoonists’ names, per the investigation’s findings.
A spokesperson for The New Yorker told Nieman Lab that Condé Nast has not authorized any large-language-model developer to train on the magazine’s cartoons. Available reporting says the 2024 content-licensing agreement between Condé Nast and OpenAI did not authorize training on The New Yorker’s cartoons.
OpenAI did not explain to Nieman Lab how its models are able to reproduce individual artists‘ signatures. Whether current safeguards block only direct style prompts or also prevent the system from reproducing identity-linked marks, such as a specific cartoonist’s pen name, remains an open question for platforms and developers to address.
