Who Made This? AI, Ownership, and the Crisis of Authorship
AI has become a sort side beat for me. I write about it mostly because it upsets me. I’ve come down hard on AI in past posts in an effort to explain to authors why it’s a problem, and why we won’t use it in any capacity for design or editorial at She Writes Press.
A couple things happened this week that brought me to a level of epiphany around the AI conversation. Kind of like leaving a couples’ therapy session and realizing you’re not even having the same conversation with your partner.
Here’s what happened: I spent some time over the holidays working on some AI education, including updating our Author Handbook and creating a video that explains our zero-tolerance AI policy. I also added a slide to the She Writes Press Orientation session that kicks off each of our new seasons in which I explained that we will not use AI for our covers, in any capacity. My voice quivered with emotion when I explained that we care about our artists, illustrators, and designers, and I saw people in this newest season nodding in the Zoom room and a few animated claps rose to the surface of my screen. I thought to myself, Okay, we can stay ahead of AI.
But then, one of my authors did the thing we ask them not to do, which I wrote about back in November (“Can You Please Make Them Stop?”) and which had prompted the addition to the handbook and the video and the slide in the first place. She sent an AI-generated image in response to our cover designs. The problem was, I wasn’t 100% sure it was AI.
The image wasn’t terrible. I thought, perhaps if the author sourced this image from an illustrator, we could ask for some changes. We could pay the illustrator and credit them, as is required in book publishing as part of licensing and copyright protocol.
So I asked her: “Is this image AI-generated?” Her response was that moment of reckoning when I realized that I’m not reaching some of my authors, ie, we are not having the same conversation.
She wrote back:
I understand your concern. Please know, the cover design I sent you is all mine and meant to be an inspiration for your cover designers. I told Chat specifically what to do. My illustration is a representation of my novel, no one else could have come up with this design, not even Chat.
This response is so sincere, but it also took the breath out of me. I’ve thought about this all week, in part because this is not the first time I’ve heard back from authors this same claim—“it’s mine”—about designs and images they’re pulling from the Internet or from AI.
A few weeks earlier, another author had chosen a cover design we all liked, but she wanted to add a landscape to the backdrop. The author did the thing of adding the landscape herself and sending it to us. The reason I ask authors not to do this is because we have to license every single image on every single one of our covers, so I asked her where she’d gotten the background image.
Her response was, “Oh, it’s all mine.”
To which I replied, “So does that mean you took the photo with your own camera and you own the image?”
“No,” she told me. She explained how she got the image from the Internet and dropped it into AI and how AI then changed the color scheme for her. She didn’t know who owned the original image, but it was so different now that it didn’t even look like the original.
I had to explain that we could not use this image. We’d have to find a similar landscape and do our own colorization. The image, in fact, was not “hers,” and had originally belonged to someone, even if we couldn’t find the source online.
I want to unpack this word and this concept: mine.
These are two very different interactions with authors over their cover art. In the first scenario, the author had a vision for her cover and asked ChatGPT to execute it. She believed that the image Chat produced for her was “hers” because she conceived of the idea that Chat delivered. In the second scenario, the author pulled an image off the Internet, dropped it into whatever AI platform, and AI altered it and in so doing created a new image.
These are two different ways we can interact with AI. AI can create/assemble/reproduce images for you, at your command, or it can take existing images and manipulate or change them.
The first example is using AI to “generate” images, while the second is using AI to alter images. In the first example, most people in the industry equate this to using stolen images. AI companies have scraped the Internet for billions of images. Billions! It’s hard to even wrap your mind around billions, but I saw a breakdown of this recently where YouTube streamer Visual Ventures shared about Sora 2, ChatGPT’s text-to-video generation model. OpenAI used 70 million short videos to train this platform. The streamer breaks this down for us here. If you were to watch 10 videos a day every day, it would take 19,000 years for you to watch 70 million videos.
Why does this matter? It matters because OpenAI did not license any of this content. OpenAI did not and has not and will not ever pay creators, the people who did the original work upon which these models are built.
Some people, I realize, do not care. I was listening to Fresh Air the other day. The episode is “A look at the ethical implications of AI.” The guest was journalist Gideon Lewis-Kraus, whose literal job here was to unpack the ethical implications of this technology in a recent piece he wrote for The New Yorker. When the interviewer, Tonya Mosley, got to the part about how AI has used the works of authors and their published works to train their models, an issue (theft, copyright violation) at the heart of a $1.5 billion class action lawsuit, Lewis-Kraus said:
“It’s not something that I tend to get all that worked up about. As the judge ruled in that case, this constitutes fair use because it’s a transformative practice. It’s not simply regurgitating stuff that it has read before; it is generalizing that stuff and then reproducing new work that follows along those lines.”
Unfortunately, this statement is inaccurate. The judge did not rule that AI training on copyrighted books is fair use. He said that training an AI model on copyrighted works could be considered a transformative use under fair use doctrine if the works were lawfully obtained. But the reason Anthropic is settling this case for $3,000 per book is because they know these works were unlawfully obtained.
To bring us toward a close here, the other thing that happened this week is that Memoir Nation hosted literary agent Alia Hanna Habib on Monday for a webinar about shopping memoir to traditional publishers. She’s a popular agent with a great Substack and a fantastic new book, Take It From Me. When the inevitable question of AI was raised in the chat during Q&A, Alia said, “If you are using AI to write your book, stay out of my fucking inbox.” She wrote about why she had this response here. In her post, she posits: “How do you expect me to advocate for you as a writer (who should be read and get paid) if you are not reading and paying other writers?”
This extends to design. Real designers are experts in composition and typography; they have an eye for what images work and why. We pay them for their expertise, and then we have to legally license those images, too (meaning pay for them), even if there are multiple images, as is the case in these two examples (see caption for details):
Human illustrators create their designs with their own hands. God help us if the future of art school is students sitting in front of computers giving voice commands to a computer and the computer spitting out the artwork. That is not any definition of fine art I ever want to be commonplace in our society. Here are a couple of examples of illustrated covers made with the minds and hands of some of our talent:
I don’t know what the future holds for us, but creative industries are concerned about AI for obvious reasons, and as writers and authors, it’s important to understand the landmine you’re walking into if you’re overly nonchalant about AI and how you’re using it. I can wrap my mind around AI being a fun tool for video and illustrations and art, especially if you’re not a trained videographer or designer or artist. All of a sudden, you have this thing at your disposal that makes you feel nearly expert.
And yet . . . I’ve been meditating on the notion of ownership—and what it means when we say “it’s mine.” “It’s mine” is an authorship claim; it’s a labor claim; and it’s an originality claim. Prompting AI to write or to design for you is not ownership. For some, maybe this doesn’t matter, but to me and countless others, especially in the publishing industry, it matters a lot.
Publishing is pushing back, too, with a lot of might. Alia Habib’s stance is hardly unique. Books with AI-generated covers have been disqualified from awards. Bookstores do not want to see anything AI-generated. My art director sent me this post from Literati Bookstore last week. Who knows what frustration prompted this, but perhaps the revelation by author Coral Hart (pseudonym) in a recent NY Times article about having published 200 AI-generated romance novels last year.
Writers and authors, we’re in a battle, whether you’re aware of it or not. When I explained copyright and reputation and ownership to both my authors in the wake of their “it’s mine” claim, both said they hadn’t realized. Both apologized. I’m not after the apologies, that’s not the point. The point is that education is half of this uphill battle. The other half is that the AI platforms have been built (and are continued to be supported) by people for whom “it’s mine” is their mindset. “It’s mine” is what led the Altmans and Zuckerbergs at the forefront of this technology to take what was not theirs to begin with—in the name of “transformative technology.” A legitimate question I have is whether transformative necessarily means good, as these companies and their stakeholders imply, and if so, good for whom?
I feel for publishers in this new world of AI. Not only must they educate authors about publishing their books with them and what all goes into it, but also battling AI's infiltration of the industry and its effect right down to the author and the making of their book. It has become a side beat that is a job in and of itself.
Such important and good points here, and the note about ownership at the end really makes me think of how sometimes, in this industry, writers want to take selective ownership of what it means to be a writer, and how, in many ways, this denudes our work of what it really is (at times hard, complex, time-consuming, etc.). Thank you for continuing to talk about AI. I find interest in it to be so disturbing, while at the same time, I am inspired to dig that much deeper into my own (real, non-AI assisted) work.