According to the US Copyright Office (USCO), an AI-generated work does not meet the “authorship” requirement of the US Copyright Act.
Last week, a U.S. federal court upheld an earlier decision by the U.S. Copyright Office to properly deny copyright protection to generative AI works submitted in the United States.
Since early 2023, nearly a dozen copyright or similar lawsuits have been filed against AI platform services that focus on whether or not AI-generated works meet the “authorship” element. Most of what we’ve seen so far concerns the training data and the relationship between the datasets it analyzes and the results it generates.
Franklin Graves of IP Watchdog listed nine cases currently focusing on these issues, ranging from Getty Images (US) and Stability AI to OpenAI, Meta and Alphabet.
Earlier this month, The New York Times updated its Terms of Service to prevent its content from being used to train machine learning systems or AI algorithms.
Unfortunately, much of the material OpenAI uses in its training datasets comes from previously established copyrighted works — without permission, credit, and compensation from the author.
The Battle of Dr. Stephen Thaler with the USCO
The last weeks Order of the Court arises from the first complaint of Dr. Thaler of June 2022 against the USCO, stating that an AI-generated work should be given copyright protection.
The work in question is an AI-generated work called ‘A Recent Entrance to Paradise’, which was the result of Dr. Thaler, ‘Creativity Machine’.
In January, he filed a motion for summary judgment arguing that under the U.S. Copyright Act, an individual should be allowed to register for copyright protection on a creative work generated by artificial intelligence.
In his motion for summary judgment, Dr. Thaler asked the U.S. District Court to issue an injunction that would require the USCO to overturn the February 2022 Review Board decision upholding the USCO’s previous positions by the copyright registry in favor of his work – and instead, re-examine his original application for copyright registration for the work.
He put forward four arguments to support his position:
- The plain language of the US Copyright Act as it currently stands allows for copyright protection of AI-generated works similar to the protection afforded to non-human entities and companies, thereby satisfying the requirement of “authorship.”
- Since the decision of the United States Supreme Court (SCOTUS) in the Burrow-Giles Lithographic Co. v. Sarony of 1800, which the Review Board argued in its opinion letter, there is a lack of knowledge and subsequently a lack of case law that supports the USCO position.
- The courts should apply the Turing Test, developed by Alan Turing in 1950, where the courts should attempt to answer the question: “Whether a machine can make something indistinguishable from a person for the purposes of copyright protection ?” In support of this, he referred to two cases of the Ninth Circuit—one involving a book created in part by spirit beings (Urantia Foundation against Maaherra) and the monkey selfie case (Naruto vs Slater).
- The work generated by AI can be classified as ‘paid work’. He stressed that while an AI is not an “employee” or an “independent contractor” under doctrine in terms of its ability to perform a contract, AI “behaves functionally” and as such should be given similar status.
The USCO did not accede to the arguments and filed a motion for summary judgment which, if granted, would automatically dismiss the case on the specific issues raised, in favor of the USCO.
Dr. Thaler submitted his reply in March, arguing that because of our advancements in technology, the current provisions of both the US Copyright Act and the US Constitution must be interpreted with outside material, including anything that predates legislation that better addresses the issue can respond. at hand. He specifically points out that “this may well be the paradigmatic case of technological evolution” that requires this kind of legal interpretation.
Where we are now
The biggest concern right now, which IPWatchdog recognizes, is that the USCO currently relies on an “honor system” to reveal whether a work has been generated by AI software or a similarly related algorithm.
This was something Dr. Thaler emphasized in his first complaint: if he had submitted the same AI-generated work, citing his company as the author, the USCO would never have known and most likely would have granted his company copyright protection. However, the Review Board did impose criminal penalties on anyone who “knowingly misrepresents a material fact” in their application for copyright registration.
The reality is that we are looking at the formation of legal precedent around copyright protection (and patent protection) for AI-generated works and the relationship between the materials the datasets are trained on and the output these machines and algorithms spit out.
Graves, delving deeper into the way forward, asked his question about where we are in this “creation-generational spectrum” that draws the line between whether a work qualifies for copyright registration or not.
And that takes us further down the rabbit hole of the conundrum we currently face with digital art and NFTs.