US appeals court rules AI generated art cannot be copyrighted

Original Article ↗ Hacker News Discussion ↗

Scope and Meaning of the Ruling

  • The case is narrow: the applicant explicitly listed the AI system as the sole author and insisted the work was “autonomously created,” while waiving any claim that he was the human author.
  • The court held that under the Copyright Act, only human authors can be initial copyright holders, so a work attributed solely to a machine cannot be registered.
  • Commenters stress this does not decide whether a human who uses AI as a tool can hold copyright; that question is left open.
  • Many compare this to the “monkey selfie” and existing Copyright Office guidance: works by animals, nature, or purely mechanical processes without human creative input are ineligible.

Human Authorship, Prompts, and Tools

  • One camp argues AI is “just a tool” like a camera, Photoshop, or spell-check: the person who sets parameters, composes prompts, and curates results should own the work.
  • Another camp notes the Copyright Office’s recent position that prompts alone are generally too indeterminate to make AI output copyrightable, especially when the same prompt yields different results.
  • Photography analogies (phone computational photography, motion-triggered cameras, tourists pressing the shutter) are used to probe where authorship really lies: setup, timing, pressing the button, or something else.
  • There’s widespread expectation that future cases will need to define how much human input (editing, masking, iterative prompting, img2img, etc.) crosses the threshold into copyrightable authorship.

Practical Implications for Art, Logos, Media, and Spam

  • Some welcome the ruling as a defense against potential AI copyright trolls (e.g., mass‑generated logo libraries used to threaten small businesses).
  • Others worry about artists using AI for “assisted” work (game art, remakes, hybrid workflows) and whether their contributions would be protected.
  • Discussion of future AI‑generated movies/games and whether users who prompt “generate a western” might hold rights, or whether this will just produce a flood of low‑value, uncopyrighted “AI sludge.”

Code, Models, and Synthetic Data

  • Many ask if AI‑generated code is similarly uncopyrightable, which would undermine proprietary claims, open‑source licenses, and “work for hire” contracts when AI assistance is heavy.
  • Disagreement over whether model weights are copyrightable: some see them as derivative of training data but transformative; others think “no human authorship” may bar protection.
  • Training on AI‑generated synthetic data raises questions about whether it can “launder” copyright or remains tainted by the original works; consensus: legally unclear.

Broader Copyright Critiques and Future Law

  • Large subthread criticizes current copyright/DMCA as primarily serving large rightsholders; proposals include much shorter terms, escalating renewal fees, or abolition.
  • Others argue copyright still protects independent creators and underpins open‑source licenses.
  • Many expect new legislation or treaty-level rules will eventually be needed specifically for AI-generated and AI‑assisted works.