Nieman Lab: ChatGPT adds real cartoonists' signatures to AI-generated images
Andrew Deck at Nieman Lab reported that ChatGPT’s image generation feature is producing cartoon images that carry the signatures of real New Yorker cartoonists — signatures the model appears to have learned from training data — without the artists’ consent. OpenAI’s tool is in effect attributing AI-generated images to specific, named illustrators.
The scale of the problem is not trivial. In the week after OpenAI expanded image generation access to 130 million users, over 700 million images were generated. Many were produced in recognizable illustrators’ styles and carry those illustrators’ signatures. An image with an artist’s signature found in a search result or shared on social media presents as that artist’s authentic work to anyone who encounters it.
Affected cartoonists have asked platforms to remove the images and are weighing legal action. The situation is distinct from the broader AI art debate about style imitation. A style is general; a signature is specific. Attribution collapses entirely when a model can attach a real person’s name to work they did not create.
For editors and publishers, the development raises a practical question: how do you verify the provenance of an image when the model generating it has learned to forge authorship signals?