ChatGPT Misattributes New Yorker Cartoon Signatures
ChatGPT’s image generator has produced New Yorker-style cartoons bearing the real signatures of more than 15 cartoonists who did not create or authorize them. OpenAI added a similarity warning after the issue was reported, but the model reportedly continued attaching real names and occasionally The New Yorker’s logo.
This is more serious than style imitation: a signature is an authenticity signal, so the failure turns aesthetic copying into false attribution. For developers, it shows why image safeguards must test rendered outputs for deceptive marks, not just screen prompts.
- –The investigation documented signatures from more than 15 real New Yorker cartoonists across repeated generations.
- –OpenAI’s prompt warning addresses similarity requests, but does not fully prevent unauthorized signatures from appearing in the final image.
- –The New Yorker says its cartoon licensing arrangements did not authorize AI training on cartoons, while contributing cartoonists retain copyright.
- –OpenAI’s own guidance recommends generic, ownable designs over imitation, underscoring the gap between policy intent and model behavior.
DISCOVERED
1h ago
2026-10-06
PUBLISHED
3h ago
2026-10-05
RELEVANCE
AUTHOR
rdmuser