The United States Copyright Office (USCO) recently issued a decision that AI-generated images cannot receive copyright protections if they are made with little human input. But the finding isn’t necessarily an indication of where US law is going to land on numerous legal questions about machine-generated works, like AI-written code, experts told IT Brew.
In September 2022, the USCO approved Zarya of the Dawn, a graphic novel by Kris Kashtanova utilizing AI-generated art, as an intellectual property protected by copyright. However, the USCO has since amended the copyright certificate to exclude the AI-generated images contained in the book. Reuters described the decision as “one of the first by a US court or agency on the scope of copyright protection for works created with AI.”
The USCO wrote that the work still qualified as a copyrightable “compilation,” finding individual elements, such as text and the selection and layout of images, are the work of the author. But it also wrote that the images themselves are “not the product of human authorship,” because Midjourney, the AI used to create them, is not a human-guided tool like a graphics editing program.
In mid-March, the USCO further clarified its position, stating that while artwork made entirely by an AI might not be copyrightable, final work can still receive protections if a human made substantial contributions, such as extensively editing a piece of AI-generated art in Photoshop.
“There’s been no doubt up to now at least that creativity meant human creativity,” Daniel Gervais, a Vanderbilt University law professor who specializes in IP and AI, told IT Brew. “Now, the machines produce something that definitely looks like it was the result of creativity, but it’s a different process entirely.”
Top insights for IT pros
From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.
“What the USCO is focusing on is that notion of creativity, and if that’s what courts base their decision on…then the USCO is probably correct,” Gervais continued.
“It’s important to stress that this is a narrow interpretation of the rules that is specific to this work, future decisions may be different,” Andres Guadamuz, a reader in intellectual property law at the University of Sussex and editor in chief of the Journal of World Intellectual Property, told IT Brew in an email. “Future decisions may change based on facts, I know of several ongoing applications to the USCO that haven’t been decided or publicized.”
“Courts can completely ignore this decision” if Kashtanova challenges the USCO determination in court, Guadamuz wrote, which her lawyers told The Register they intend to do.
Guadamuz stressed the underlying legal questions in the USCO case are different from those in other high-profile fights over AI and copyright, like the proposedclass-action lawsuit against Microsoft and GitHub’s CoPilot. The assistant’s developers trained it on open-source repositories, including GitHub, claiming fair use, and it is now an enterprise-grade commercial product.
The dispute is over whether CoPilot, which does not cite the original projects when generating new code, violates the terms of open-source licenses that require attribution on reuse and CoPilot critics say aren’t public domain.
“This is going to be centered on the narrow interpretation of what is a derivative, and whether the terms of existing licenses are infringed by GitHub,” Guadamuz wrote.—TM