At 4:01 p.m. on Monday, Nieman Lab published Andrew Deck’s report that ChatGPT, asked for a New Yorker-style cartoon, will produce one — and then sign it with the name of a working New Yorker cartoonist. Brendan Loper found out the way everyone finds out these days: somebody sent him a screenshot of his own signature under a drawing he never made. “I’m not a territorial person,” he told Deck, “but my name is my name. It felt very much like a violation of my personhood.”
The Hacker News thread on the piece had 341 points and 251 comments by Tuesday morning, and it split the usual way. One camp called it theft. The other camp recited the doctrine: style isn’t copyrightable, the model learned from the corpus, cope. Both camps are arguing about the picture. The picture isn’t the asset. The signature is.
The Signature Was Never Art. It Was a Credential.
Consider what a cartoonist’s signature actually does. It is nine letters of ink that carry no aesthetic information whatsoever. Strip it off and the drawing is unchanged. What it carries is a claim: that a named person, working under the standards of a specific institution, made this and stands behind it.
That is not expression. That is attribution. Which means it lands in an entirely different body of law than the one the AI industry has spent three years fortifying against. Fair use is a defense to copying. It is not a defense to impersonation. The Lanham Act has handled false endorsement and false designation of origin since 1946, and most states have right-of-publicity statutes, several of them with statutory damages that don’t require the plaintiff to prove a dollar of lost income. You do not have to show that a reader was confused into paying Loper’s fee to somebody else. You have to show that a name was used to sell something the name didn’t make.
“Every fair-use brief in this country is about copying,” said one plaintiffs’-side lawyer working the right-of-publicity docket, waiting out a status conference in a hallway off the ninth floor of the federal building in San Francisco. “There is no fair-use defense for signing a name. That’s not a copyright question. That’s a lie with a byline.”
A Fake Presidential Video Gets Fact-Checked. A Fake Gag Cartoon Doesn’t.
Here is why this particular forgery is a bigger deal than it looks, and smaller than the alarmists want.
The deepfake of a president is contested territory. Networks flag it, platforms label it, researchers authenticate it, and the whole apparatus of verification swings into gear because the stakes are enormous and legible. A counterfeit New Yorker cartoon has none of that. Nobody’s livelihood turns on whether the guy at the next desk over actually drew it. No editor’s job depends on catching it. The cost of checking is nonzero and the benefit of checking is roughly zero, so nobody checks.
That is the actual news in Deck’s report, and it isn’t about art. The forgery economy has moved down-market, into the enormous unguarded middle where verification is nobody’s job. And notice what the user is buying. Not a drawing — anybody can get a drawing. They’re buying a drawing that appears to have been vetted by a magazine. The machine’s real product was never talent. It’s borrowed institutional adjacency, sold by the syllable.
The entire “what does AI mean for human creativity” argument has been a decade-long distraction from that. The scarce good was never the picture. It was the credential.
The Plaintiff Who Wins This Won’t Be the Cartoonist
Now the uncomfortable part, for readers inclined to cheer Loper on.
He has a real claim. He also has a small one, against a defendant whose market capitalization starts with a number most people can’t picture, in a forum that will be chosen by the other side’s motion. Cartooning is not a lucrative trade. Discovery in a federal case is. A single artist suing over a forged signature is a moral victory with a legal bill attached, and everyone in the room knows it.
But the signature was not the only thing counterfeited. The house style was. The New Yorker — a Condé Nast property, with a trademark portfolio, a licensing business, and a legal department that does not need to run a GoFundMe — has a claim that is bigger, cleaner, and better funded than Loper’s. When the dust settles, the entity most likely to extract a settlement and a filtering commitment from OpenAI is not a cartoonist. It’s the magazine.
Which means the artists end up as exhibits in somebody else’s case. If you care about the person whose name got borrowed, sit with that for a second: the instrument that protects him is the instrument that absorbs him. Copyright reform would take an act of Congress. Trademark law already exists, already works, and already belongs to the people with the biggest brands.
And the remedy is cheap, which is the part that ought to annoy everyone. A model that declines to render a living artist’s signature is not a hard engineering problem. It’s a list and a classifier. A system that will not help you pick a lock will happily help you sign a stranger’s name to a picture he didn’t draw, because nothing in the pipeline was asked to care. That isn’t a capability limit. It’s a priority.
No new statute is required for that. No Brussels framework, no Senate hearing, no blue-ribbon commission on the future of illustration. Just an existing cause of action, a name that belongs to somebody, and a company that would rather filter a string of characters than explain to a jury why it didn’t.
Sources
- cartooning | Nieman Journalism Lab
- David Forrest on X: “#OpenAI’s chatbot isn’t just imitating the New Yorker’s style. It’s also falsely attributing #AI-generated images to real artists.” / X
- Nieman Lab on X: “ChatGPT is adding real cartoonists’ signatures to fake New Yorker cartoons https://t.co/i5ZxODkVnp” / X
- ChatGPT is adding real cartoonists’ signatures to fake New Yorker cartoons | Nieman Journalism Lab
- Nieman Journalism Lab
- Andrew Deck | Nieman Journalism Lab