Mashable: Who owns a story written by ChatGPT?

Every month, Clarkesworld — a monthly science fiction and fantasy magazine — receives about 1,100 submissions from writers hoping to get published. To prepare each issue, Publisher and Editor-in-Chief Neil Clarke and his staff read through the pieces and narrow down their selections for publication.

But over the past few weeks something changed: Side hustlers started using ChatGPT. Blogs highlighting gigs to supplement other work, fueled by the ever-present hunt for clicks, began claiming you could make easy money with ChatGPT by asking the AI-generated chatbot to write a short story, copying the text it produces, and submitting that work to literary magazines that pay published writers.

These side hustle sites have broadcasted lists of outlets that pay writers for published submissions, like Clarkesworld, and encouraged people to submit work there for years — the difference now, though, is that the sites are encouraging readers to use AI tools like ChatGPT to generate the stories. It takes far less time to instruct a chatbot to produce content than it does to actually write that story, and hundreds of people have taken advantage of the opportunity to potentially make a quick buck. 

In February alone, Clarke said the magazine received 700 submissions written by humans and 500 written by AI. Despite the fact that the AI stories are pretty easy to spot — they're often written poorly and frequently even have the same titles — they overwhelmed the submission system for Clarkesworld so thoroughly that Clarke was forced to close submissions.

"One of the things I'm comparing it to is spam," Clarke told Mashable. "So right now we're in pre-spam filter days, but with current-day spam levels. So we need to find some ways to knock this down."

As spam tends to work, Clarkesworld isn't the only literary magazine to be hit. The Magazine of Fantasy & Science Fiction(Opens in a new tab) and Asimov’s Science Fiction(Opens in a new tab) also reportedly have been flooded by submissions generated by AI chatbots.

"I knew it was coming on down the pike, just not at the rate it hit us," Sheree Renée Thomas, the editor of The Magazine of Fantasy & Science Fiction, told the New York Times(Opens in a new tab).

Clarkesworld isn't particularly interested in publishing works of fiction written by a computer, not only because the works are usually of poor quality, but also because this aspect of the rapid rise of AI access has become ethically and morally blurry, with some raising red flags of straight-up plagiarism.

Who owns the work?
Beyond overwhelming systems for submissions, this new AI tool has threatened to upend the literary world as we know it. AI isn't new to the field: Writers have used AI-based writing assistants like Sudowrite(Opens in a new tab), Jasper(Opens in a new tab), Writesonic(Opens in a new tab), and more for years. And there are plenty of ways to use these and ChatGPT as genuinely helpful, ethical writing tools. These resources can help with plot holes, naming characters, creating an outline, ironing out dialogue, and more. But they're are a bit clunkier to use than ChatGPT, and they aren't marketed to the public in the same way. 

"Unless [ChatGPT] is constrained, I hate to say this, but it will have a negative impact on the writing profession without a doubt and on the quality of what gets published," said Mary Rasenberger, CEO of the Authors Guild, the nation’s oldest and largest professional organization for published writers. 

This is a problem not limited to authors and publishers. 

"As people say, books are the arteries of democracy; ideas spring from books," Rasenberger told Mashable. "We need to protect book culture, but how do you do that in a world that is flooded with AI books?"

 We want our arts to be human artists reflecting their current experiences and hopes and fears about our current world.- Mary Rasenberger, CEO of the Authors Guild

Because the model for ChatGPT is trained on a body of created works, "it is still unclear what the legal precedent may be for reuse of this content, if it was derived from the intellectual property of others," Bern Elliot, an analyst at Gartner, said in a blog post(Opens in a new tab). 

ChatGPT doesn't include citations or attributions to original sources, and it’s unclear who can copyright or claim ownership of AI-generated works: the person who requested something from the AI, OpenAI itself, or someone else? Determining this answer will be imperative to clarifying ChatGPT's role in the literary world.

Credit where credit is due
Precedent already is underway in the industry. After a graphic novel, Zarya of the Dawn, used images created by AI technology, the U.S. Copyright Office decided that the images wouldn't be protected by copyright(Opens in a new tab), because copyright can only be granted to works created by a human being. Beyond copyright, the Authors Guild is encouraging Congress to pass legislation that says the works that ChatGPT is trained on have to be licensed.

"We're not asking any AI developer to go get rights from every writer in the world, much less everyone who posts a blog online," Rasenberger said. "But at least for books and for journalism, to pay into a collective licensing system [which] would then pay out sums of money to the creators."

That kind of program would ensure that at least some money goes back into the pockets of creators so that they can continue to create and writers can continue to write.

"We're not saying, 'Oh, we're going to stop AI in its tracks,'" Rasenberger said. "We're not going to do that because AI can be useful. But you have to recognize that if we don't find a way to compensate writers, in this case, the case of text for training AI on their works, we're going to put writers out of business."

AI isn't inherently insidious
Despite all the trouble this has caused Clarke and others, he actually doesn't completely blame the AI itself. Instead, he argues, the people using the tools are the ones making his life more difficult.

"This is really a human problem," Clarke said. He added that while the AI is problematic, engineers should have incorporated things like watermarking long before they released the program as is. "It's still a very human province — the humans are burying me. It's not an AI submitting stories, which maybe someday it'll do on its own, and then I'll be in a new fresh hell. But at this point, it's humans."

That also means that AI isn't anywhere near replacing human writers.

"It has no emotion, has no ability to think, no ability to generate any new ideas," Rasenberger said. "We don't want our arts to be limited to that. We want our arts to be human artists reflecting their current experiences and hopes and fears about our current world. That's how arts have grounded civilization since the beginning of human civilization. And I fear that we will lose that if we do not tread into this AI world very carefully. "

Humans will have to be the ones to ensure the safety of the arts — with legislative solutions, copyright law, and more. But we're not at risk of replacement just yet. Sheila Williams, the editor of Asimov’s Science Fiction magazine, told The Verge(Opens in a new tab) that she wants writers to know she isn't going to miss their work because she's "inundated with junk," offering reassurance that "the mind that crafts the interesting story is not in any danger."

ChatGPT and AI are bound to change the way we brainstorm, write, edit, publish, and create. But, as long as we treat this transition with care, it doesn't have to come at the expense of respect for human talent.