Copyright does not protect AI-generated content in EU

Original Article ↗ Hacker News Discussion ↗

EU copyright law is being interpreted to deny protection to works generated entirely by AI, putting such content effectively into the public domain while still allowing underlying human-authored source material to remain protected. Commenters explore how this interacts with translations, software code, logos, and mixed human–AI workflows, and note parallels and differences with emerging US case law. Many see unresolved problems around how to prove human contribution, how watermarks or detectors might be used in court, and what this means for open source licensing, patents, and future incentives to create.

Scope of the EU Ruling

  • Content “entirely generated” by AI is not protected by copyright in the EU.
  • Many see this as aligned with prior principles: copyright requires an “intellectual creation” by a human author.
  • Several commenters say this outcome was expected; others note it’s still significant that it’s now explicit.

Derivative Works & Translations

  • Consensus: lack of copyright in AI output does not remove underlying copyrights.
    • Example: AI translation of a copyrighted book is still restricted by the original’s copyright, even if the translation itself isn’t protectable.
  • For public-domain works, an AI translation would not create a new copyright layer, unlike a human translation.

Human Contribution Threshold

  • General view: mere prompting or choosing among AI variants is not enough for copyright.
  • Debate over where the line is:
    • Minor edits likely only protect the edits, not the full AI output.
    • Heavier transformation could create a new human-authored work, but “how much is enough” is unresolved and seen as case-by-case.
  • Some argue agentic workflows and extensive human steering should qualify; others say if you “didn’t write it, it’s not yours.”

Software, Open Source, and Licensing

  • Question whether AI-generated code can be copyrighted or licensed under MIT/GPL/BSD.
    • Common view: licenses can’t be enforced on uncopyrightable AI portions, but can still apply to surrounding human-written code.
  • Concern that watermarking (e.g., in code) could later be used to challenge copyright claims on mixed human/AI work.
  • Some fear large products using AI-generated code (OSes, office suites) might face copyright uncertainty.

Enforcement, Proof, and Detection

  • Many note practical difficulty: proving something was AI-generated versus human-authored is hard; watermarking and detectors are imperfect and raise false-positive concerns.
  • Some foresee new industries or legal tests around evidencing human contribution.

Normative and Policy Views

  • Split opinions:
    • Some celebrate AI outputs defaulting to public domain and call for broader copyright rollback.
    • Others worry that undermining IP protection (especially where AI is used heavily) will disincentivize investment and harm creators.
  • Ethical disputes over training on copyrighted data without consent recur, with analogies to human learning both defended and criticized.