Summer must be coming, because the courts are starting to heat up with copyright decisions in artificial intelligence (AI) cases.
We’ve previously written here, here, and here about Dr. Stephen Thaler’s attempts to register copyright in “A Recent Entrance to Paradise,” a work created by his AI platform DABUS. On March 18, 2025, in Thaler v. Perlmutter, the U.S. Court of Appeals for the District of Columbia Circuit held, as the Copyright Office and the district court below had, that only humans can be “authors” under the Copyright Act, effectively closing the door on Dr. Thaler’s registration efforts.
As the D.C. Court of Appeals explained, the Copyright Act contains numerous statutory indicators that “authors” for purposes of copyright are humans, not machines. For instance:
- Copyright “vests initially in the author,” meaning the author immediately gains “exclusive rights” to their work upon the work’s creation. “Because a copyright is fundamentally a property right created by Congress, and Congress specified that authors immediately own their copyrights, an entity that cannot own property cannot be an author under the statute.”
- The duration of a copyright is limited to “the author’s lifespan or to a period that approximates how long a human might live,” plus a limited time thereafter. “Of course, machines do not have ‘lives’ nor is the length of their operability generally measured in the same terms as a human life.”
- Copyrights are inheritable by widows and widowers, children, and grandchildren. “Machines, needless to say have no surviving spouses or heirs.”
- Copyright transfer requires a signature. “Machines lack signatures, as well as the legal capacity to provide an authenticating signature.”
- Authors of unpublished works are protected regardless of their nationality or domicile. “Machines do not have domiciles, nor do they have a national identity.”
- “[A]uthors have intentions,” as joint works are prepared by two or more authors who intend their separate contributions to be merged. “Machines lack minds and do not intend anything.”
- All discussion of machines in the Copyright Act “indicates that machines are tools, not authors.”
Drawing on these contextual cues, the D.C. Court of Appeals concluded that “[t]he human-authorship requirement, in short, eliminates the need to pound a square peg into a textual round hole by attributing unprecedented and mismatched meanings to common words in the Copyright Act.” Thus, “the current Copyright Act’s text, taken as a whole, is best read as making humanity a necessary condition for authorship under the Copyright Act.”
This latest decision in the Thaler saga—which may not be the last word, as Dr. Thaler’s counsel has indicated he will appeal to the U.S. Supreme Court—not only has ramifications for authorship of AI created works but also may impact future infringement claims against those who use generative AI to create output. One could envision that defendants who use AI to generate allegedly infringing works would simply blame the AI and argue they cannot be held liable for what the platform generated since, as Thaler recognizes, only humans can create copyrightable works.
But as the D.C. Court of Appeals recognized in Thaler, it is possible to register copyright in AI-generated works where sufficient human intervention, such as significant and creative prompts, led to the generation of the work. The Copyright Office issued guidance to that effect in 2023. Dating back to the 1970s Congress grappled with whether machines (at that time, computers, and previously cameras and other once-novel machines) could be copyright authors and concluded that they could not—but Congress never concluded that works created by humans using machines are not copyrightable (and they are, in fact, copyrightable). As the Supreme Court recognized decades ago, the requisite creativity for copyrightability is quite low, holding the arrangement of a phonebook (remember those?) was copyrightable. Copyright owners will push for courts to find even minimal human effort to generate a work using AI—say a handful of user prompts—is sufficient to establish infringement liability over the person who generated the accused work using AI.
Recent district court decisions show that infringement claims based on using copyrighted works to train AI and directed to similarity in output generated by AI are viable, at least at the pleadings stage, and at least against the owners of the implicated AI platforms. For instance, on March 26, 2025, in The New York Times Co. v. Microsoft Corp., the District Court for the Southern District of New York denied OpenAI’s motion to dismiss direct and indirect copyright infringement claims brought by The New York Times, finding the Times sufficiently alleged direct infringement by OpenAI’s use of its copyrighted works to train OpenAI and indirect infringement by OpenAI for works subsequently generated by OpenAI users.
Also on March 26, 2025, however, in Concord Music Group, Inc. v. Anthropic PBS, the District Court for the Northern District of California granted a motion to dismiss indirect infringement claims brought by music publishers against Anthropic for its AI platform Claude, which was allegedly trained on the publishers’ copyrighted song lyrics and allegedly generates infringing lyrics as output. That decision, however, seems to be of limited effect; the motion was granted because the publishers failed to sufficiently allege any instances of direct infringement by the platform’s users, a predicate for an indirect infringement claim, a seemingly readily correctable omission. Indeed, the publishers have indicated they will re-plead. The parties seem to have already themselves realized the potential for direct infringement by Claude’s users, having entered an agreement in December 2024 putting “guardrails” in place to prevent Claude’s output from infringing the publishers’ works going forward.
With a number of AI cases pending for a few years now, 2025 promises to be an interesting year for AI-and-copyright. Ultimately Congressional intervention will almost certainly be needed to iron out the myriad of issues, but until then, as the courts grapple with these issues, we will continue to report on new developments.
Example of a work “created” by the author via Copilot, using approximately three prompts. Courts will be tackling copyright and AI issues for some time to come.
This article is provided for informational purposes only—it does not constitute legal advice and does not create an attorney-client relationship between the firm and the reader. Readers should consult legal counsel before taking action relating to the subject matter of this article.
