Copyright Is the Latest Battle in the War Over A.I.

Increasingly sophisticated artificial intelligence (AI) is penetrating the mainstream, and no one is quite sure what to make of it. Students protested after university administration used the AI ​​text generator ChatGPT to compose a statement following a mass shooting. Systems like Midjourney can create artworks that are so sophisticated they can win prizes, much to the chagrin of flesh-and-blood artists.

Now the US Copyright Office (USCO) has gotten into the debate, stating that images produced by AI are not copyrighted.

In 2022, artist Kristina Kashtanova published the graphic novel Zarya of the Dawn free on the internet. Kashtanova wrote all of the text and arranged the pages for the book, but used Midjourney to create all of the visual images, adding text prompts to it. They applied for a copyright on the work, which USCO granted in September 2022, specifically “to set a precedent.”

But the following month, USCO changed its position due to Midjourney’s role, stating, “Copyright under US law requires human authorship.

In a decision released last week, USCO ruled that Kashtanova can copyright the comic’s text and the “selection and arrangement of images and text,” but not the images themselves.

The case raises the question of what exactly constitutes an original creation when an AI is involved, and who owns it. Last year, VentureBeat asked “Who owns” the images created by OpenAI’s DALL-E, another AI image generator. Legal experts and even OpenAI itself seemed uncertain.

Typically, in a work-for-hire situation, whoever commissions a work owns the copyright. And in a decision last year, USCO concluded that an AI couldn’t copyright its own art anyway “because copyright…requires human authorship.”

To be copyrightable under US law, a work must exhibit “at least a minimum level” of creativity, a standard so low it includes almost everything but the phone book. The 9th Circuit Court of Appeals ruled in 1997 Urantia Foundation against Maaherra that “some element of human creativity must have taken place for the [work] be copyrighted.”

When USCO denied Kashtanova a copyright to the comic’s images, she cited the US copyright code against “register[ing] Works made by a machine or a purely mechanical process that occurs randomly or automatically without creative input, or Intervention of a human author.“But as the decision sums up, create Zarya of the Dawn required a tedious trial and error process. Kashtanova may not have illustrated the images, but they fed the machine “hundreds or thousands of descriptive prompts” and trawled through “hundreds of iterations” of results to find the perfect image for each use.

Regarding the Midjourney process, the decision states “The The initial prompt from a user generates four different images based on Midjourney’s training data. While additional prompts applied to one of those first images may affect the subsequent one Images, the process is not controlled by the user as it is unpredictable what Midjourney is created in advance.“

Van Lindberg, Kashtanova’s attorney, wrote over the weekend that USCO used the wrong legal standard: “The standard is whether there is a ‘minimal level of creativity,’ not whether Kashtanova ‘could predict what is halfway [would] create in advance.’ In other words, the Office wrongly focuses on the output of the tool rather than that entrance from humans.”

There is precedent for the Kashtanova case: Michael Noll, a researcher at Bell Labs, successfully copyrighted an AI-created image in 1965, on the premise that the machine created the image using an algorithm written by Noll. After Kashtanova’s first copyright was granted, Benj Edwards enrolled Ars Technica that “this is the first time we know that an artist has filed a copyright for art” created by the types of AI currently under discussion.

The USCO decision reflects the current capricious debate about artificial intelligence. Is AI a threat to the established order, or is it just another tool for artists to learn and master?