ChatGPT Puts Real Cartoonists' Signatures on Fake Cartoons

ChatGPT is generating New Yorker-style cartoons that carry the signatures of real cartoonists from the magazine, according to an investigation Nieman Lab published on October 5, 2026. The outlet documented more than 15 signatures. It matters to those artists, and to any business that puts AI-generated images on its website.

What we know

Who and when: Nieman Lab published the investigation, by Andrew Deck, on October 5, 2026. It is a news outlet's reporting, not an OpenAI announcement.
How many signatures: Deck says he documented more than 15 New Yorker cartoonists whose signatures were used by OpenAI's image generator without permission or compensation. He names Brendan Loper, Emily Flake, Harry Bliss, Joe Dator and Pat Byrnes, among others.
The best-known case: a cartoon that spread in late August 2026 carried the signature "BLOPER", Loper's pen name, in the bottom right corner. He did not draw it. One tweet of the image received 25,000 likes. The person who first posted it wrote that she had only asked ChatGPT for a New Yorker-style cartoon, according to Nieman Lab.
It does not always happen: in the reporter's tests, not every image came with a signature and some carried gibberish names. The signatures that did appear were regularly those of real contributors, the outlet says.
What OpenAI said: a spokesperson told the outlet the company appreciates people flagging bugs and unintended model behavior so it can address them. It declined to answer how the signatures might have got into its models, according to Nieman Lab.
What happened next: after the reporter contacted OpenAI, ChatGPT began showing a message about its limits on similarity to third-party content when asked for cartoons in that style. As of publication it was still signing some with real names, the outlet says.
What the magazine said: a New Yorker spokesperson told the outlet that Condé Nast, its parent company, has never allowed a language-model developer to train on its cartoons. Some images also reproduced the magazine's logo, which its deals do not allow either.
The legal view, from one expert: James Grimmelmann, a professor at Cornell Law School, told the outlet that ChatGPT is mimicking a general style rather than copying a specific cartoon, which is the main hurdle for a copyright suit. He sees another possible route in "right of publicity" law, which would require proof of commercial use.

What changes and what doesn't

No product and no rule has changed: OpenAI has announced nothing. Its usage policies, effective since October 29, 2025, already bar attempts to infringe other people's intellectual property, as well as impersonation. Those rules are about what a user does. The investigation describes something else: signatures that show up without anyone asking for them.

What changes is the level of care: a generated image can carry someone else's signature or logo even when the prompt is generic.

The investigation has limits. It does not say how often signatures appear, how many prompts were tested, or which model was used. It does not establish how the system learned them: the outlet raises scraping from the internet as a possibility. It mentions no lawsuit by these cartoonists over the signatures. And its author says he found no evidence that the cartoons are being bought or sold.

How to tell if it affects you

It affects you if you publish AI-generated images on your website, your store or your ads. Seven checks before you upload one:

Look at the corners at full size. In the viral case the signature sat in the bottom right. Zoom to 100% and look for strokes shaped like letters or a flourish. They do not show in a thumbnail.
Look for logos and mastheads. The New Yorker spokesperson described images carrying the magazine's logo. Check signs, labels and stamps inside the scene too.
Do not prompt "in the style of" a named magazine, brand or artist. That is how the viral case began. Describe what you want in your own words: technique, colors, framing.
If a signature or a logo appears, discard the image and generate another. Erasing or cropping it does not change whose work the drawing resembles.
Run a reverse image search before using the image on a cover or in an ad. If it returns near-identical works, do not use it.
Keep a record: tool, date, exact prompt and who reviewed the image.
Separate a meme from a business website. The expert quoted is describing United States law, and names commercial use as the condition. A company's website or ad looks more like that case than a joke shared on social media. That is this article's reading, not a legal test.

This article is not legal advice: the rules on copyright and on the use of a person's name differ from country to country.

Sources

Updates: any explanation from OpenAI of where the signatures came from, or any change to its image generator, will be added here.