The question
Where the law stands
Copyright protects human creativity. A work with no human author gets no protection.
Typing instructions isn’t authorship. Selecting, arranging, and editing can be.
Dozens of lawsuits ask whether copying works to train a model is infringement.
Transformative purpose helps. Pirated sources and market harm hurt.
Publishers, archives, and labels are signing data deals instead of waiting for verdicts.
D.C. Circuit · 2025
No human author, no copyright.
Stephen Thaler listed his AI system as the sole author of an image. The Copyright Office refused, a district court agreed in 2023, and the D.C. Circuit affirmed in March 2025: the Copyright Act requires a human author.
U.S. Copyright Office · 2023
The story counts. The Midjourney images don’t.
A comic book made with Midjourney kept its registration for the human-written text and the selection and arrangement of panels, but not for the individual AI-generated images.
D. Delaware · 2025
First court to reject fair use for AI training.
Ross used Westlaw headnotes to build a competing legal-research tool. The court held this was not fair use, a ruling about a non-generative AI that is now on appeal.
N.D. California · 2025
Training was fair use. Pirated books weren’t. $1.5B.
Judge Alsup found training on lawfully acquired books transformative, but not building a library from pirate sites. The case settled for $1.5 billion, roughly $3,000 per work.
N.D. California · 2025
Meta wins this round. Market harm is the battleground.
The court ruled for Meta because the authors hadn’t proven market harm, while warning that AI flooding the market with substitutes could defeat fair use in other cases.
S.D. New York · 2023–
The newsroom versus the model.
The New York Times alleges millions of its articles were used without permission to build products that compete with it. The case is still in litigation.
UK High Court · 2025
A model is not a copy.
In the UK case, the court held that Stable Diffusion’s trained weights are not an “infringing copy” of Getty’s photos. Getty won only narrow trademark points on watermarks.
N.D. California · 2023–
Artists versus image generators.
Visual artists sued Stability AI, Midjourney, and DeviantArt. In 2024 their core copyright claims survived dismissal, and the case is still being litigated.
In Burrow-Giles v. Sarony, the Supreme Court upheld copyright in a photograph of Oscar Wilde. The machine took the picture, but the photographer’s choices of pose, light, and costume made it authorship.
The Register of Copyrights’ annual report flags works “produced by computers” and asks the question we’re still asking: was the computer an assisting instrument, or did it conceive the work?
Congress’s CONTU commission concludes computers are tools like cameras or typewriters, and that authorship belongs to the people who use them.
In Naruto v. Slater, the Ninth Circuit holds that a macaque can’t sue for copyright in a selfie it took. The Copyright Office’s rules already said it wouldn’t register works made by nature, animals, or machines without human input.
The Copyright Office issues registration guidance and narrows Zarya of the Dawn. A court rejects Thaler’s AI-authored image. Artists, Getty Images, authors, and the New York Times file the first wave of training lawsuits.
The Copyright Office publishes its reports on copyrightability and on training. The D.C. Circuit affirms Thaler. Ross rejects fair use, while Bartz and Kadrey find training transformative. Anthropic settles for $1.5B, and the EU AI Act’s copyright duties for general-purpose models take effect.
Appeals, new lawsuits, and a growing licensing market keep redrawing the line between tool and author, and between learning and copying. The question is settled for no one.
Get in touch
Partnerships, press, research, or anything else on your mind.
Received
We’ll reply to the email you gave us.