COURT-RULING

Thaler v Perlmutter

Updated 2 May 2025
court-rulingcopyrightauthorshipusa

Thaler v Perlmutter

Can AI be an author? The US courts said no.

Stephen Thaler applied for a copyright on an image generated entirely by his AI system, DABUS (Device for the Autonomous Bootstrapping of Unified Sentience). He listed DABUS as the author. The US Copyright Office refused. He sued. He lost.

The ruling is simple but consequential: copyright protection requires human authorship.

The Facts

Thaler created an AI system called DABUS that autonomously generated an artwork called “A Recent Entrance to Paradise.” Thaler sought to register the copyright with the US Copyright Office, listing DABUS as the author and claiming ownership himself as the owner of the AI.

The Copyright Office denied the registration. Thaler challenged the denial in court.

The Ruling

Judge Beryl Howell ruled clearly:

“Copyright has never stretched so far as to protect works generated by new forms of technology operating absent any guiding human hand.”

The court held that:

  • US copyright law requires human authorship
  • An AI system cannot be an “author” under the Copyright Act
  • The longstanding requirement of human creativity is a constitutional and statutory principle
  • The Copyright Office correctly denied registration

Why It Matters

What it decides: Works generated entirely by AI, without meaningful human creative input, are not copyrightable in the US. They enter the public domain.

What it doesn’t decide: The harder question — what about AI-assisted creation where a human directs the AI? The court explicitly left this open. If you write a detailed prompt, curate the output, and make creative selections, is the result copyrightable? That’s still unresolved.

Implications:

  • Pure AI output = public domain (in the US)
  • AI-assisted output = case by case, unresolved
  • This creates strange economics: you can generate an image with AI, but your competitor can copy it freely
  • The Copyright Office has since issued guidance: it considers AI-generated elements on a case-by-case basis, denying protection for AI-generated portions while allowing it for human-authored portions

The International Picture

Other jurisdictions are reaching different conclusions:

  • UK — Copyright law may allow computer-generated works to be owned by the “person who made the arrangements” for creation. Untested with modern AI.
  • EU — Generally follows the originality/human authorship approach, but member state implementations vary
  • China — Beijing court ruled in 2023 that AI-generated images can be copyrighted if the human’s creative input was sufficient
  • Japan — Broadly permissive on AI training, but authorship questions similar to US

This patchwork means an AI-generated work might be copyrighted in one jurisdiction and public domain in another. The inconsistency is a real problem for global businesses.

Connection to Training Data Cases

This ruling sits alongside NYT v OpenAI and Getty v Stability AI as the three pillars of the AI copyright debate:

  1. Can you train AI on copyrighted data? → NYT v OpenAI, Getty v Stability AI
  2. Can AI create copyrighted output? → Thaler v Perlmutter (answer: no, not without human authorship)

Together, they define the input and output of AI copyright law.

Go Deeper

Sources

enes