Copyright does not protect AI-generated content in EU
mathstodon.xyz
August 21, 2026
1 min read
55/100
Summary
Mastodon’s web application requires JavaScript to be enabled. Users who cannot or do not want to enable JavaScript can use a native Mastodon app for their platform instead.
What the discussion said
Commenters largely accept the core principle that a work produced with no meaningful human authorship should not automatically create a new monopoly. Several readers see that as a sensible extension of copyright’s human-creator foundation: AI can generate useful material, but pressing a model into service is not obviously equivalent to writing, illustrating, or programming the result. They also note that this does not make AI output consequence-free. A generated translation, game remake, or codebase can still infringe the rights in its source material, even if the new AI-produced layer earns no separate protection. The real argument is over the messy middle, not fully autonomous output. Readers worry that prompt iteration, agent steering, selection among candidates, and light editing will force courts to invent arbitrary tests that lag behind actual human-AI workflows. The practical stakes are clearest for open-source licensing and commercial R&D: purely generated code may be usable but cannot reliably carry GPL, MIT, or similar restrictions, while companies may retreat to secrecy if AI-assisted inventions cannot be protected. Some regard that loss of exclusivity as healthy public-domain expansion; others say it undermines incentives. A further undercurrent holds that outputs should not be privileged while models remain trained on creators’ work without consent or payment.
Where opinion split
The sharp dispute is whether detailed prompting and iterative steering amount to authorship. One side argues that a person who does not directly create the expression cannot claim it merely by directing a statistical system; the other says real production work is rarely wholly autonomous, and excluding guided AI workflows would create an artificial, unworkable standard. Both sides agree that trivial cosmetic edits should not magically monopolize an otherwise unprotected output.
Community Sentiment
Positives
- Keeping wholly autonomous AI output outside copyright prevents a model operator from receiving exclusive rights merely for running a generator.
- The ruling can enlarge the practical public domain, letting people reuse genuinely AI-only material rather than locking routine generated output behind new monopolies.
- Separating human contributions from model contributions preserves copyright for real writing, editing, curation, and design work without pretending the machine is an author.
- Readers stress that losing copyright does not erase source-rights claims, so AI-generated translations or remasters remain constrained by the original work’s copyright.
Concerns
- No convincing boundary exists yet for iterative prompting, agent supervision, selection, and revision, leaving AI-assisted creators unable to predict what they can protect.
- Open-source projects may struggle to enforce GPL, MIT, BSD, or Creative Commons terms over AI-written portions, weakening the reciprocity or attribution rules maintainers rely on.
- If AI-created technical advances cannot be protected through copyright or related IP, firms may lock methods away as trade secrets instead of sharing them through patent-like disclosure.
- Several commenters see a moral asymmetry in denying rights to AI output while models may have absorbed creators’ work without meaningful consent or compensation.
