Report · pluralistic.net
Copyright Office is restoring copyright to something you have to work to get
pluralistic.net
Source published October 31, 2026
Summary: 1 min read
45/100
Summary
The U.S. Copyright Office has repeatedly held that copyright protects human authorship and does not extend to material generated solely by AI systems. Under that position, human contributions to AI-assisted material—such as creative writing, editing, or selection—may be protectable, while the AI-generated portions are not. Cory Doctorow says this distinction could require companies to document which elements of commercial works were created or altered by people and which were generated by AI. U.S. copyright once required formalities including deposit with the Library of Congress, registration identifying authorship claims, and copyright notice. The 1976 Copyright Act eliminated most formalities as the United States joined the Berne Convention, although registration remains necessary to obtain statutory damages of up to $150,000 per work. Copyright now generally arises when an original work is fixed in a tangible medium, and unregistered works can remain protected for the author’s life plus 70 years. Doctorow argues that the lack of a comprehensive registry has made it difficult to identify copyright owners and determine the status of many works. He predicts that businesses using AI will recreate record-keeping requirements privately in order to establish and defend copyright claims over human-authored portions of their products.
Key Takeaways
- The U.S. Copyright Office’s position is that only human-authored works qualify for copyright protection.
- Human creative edits, writing, and selection within AI-assisted material may be copyrighted even when purely AI-generated elements are not.
- U.S. copyright formalities were largely abolished by the 1976 Copyright Act, but registration is still required for statutory damages.
- Doctorow predicts that companies using AI will need detailed records identifying AI-generated and human-created elements to support copyright claims.
What the discussion said
Commenters mostly treated the story as a clarification of a long-standing rule, not a sudden Copyright Office revolt: US copyright attaches to human authorship and creative choices, not effort alone. Several readers argued that this is a useful constraint on firms hoping to replace artists wholesale with generated material, because unedited model output cannot become a proprietary asset merely because a company spent heavily producing it. They also noted a practical middle ground: humans can still claim protection for their own selection, editing, arrangement, and other demonstrable contributions around AI-assisted work. The thread repeatedly separated that question from the far messier dispute over training. One early claim that a coding lawsuit had recognized AI as an author was corrected; the cited decision concerned whether outputs were copies of training code, not whether a model can hold copyright. That distinction mattered because participants saw output ownership and input liability as legally independent, even if they fear industry will push for a regime that protects generated products while excusing mass ingestion of copyrighted work. Readers also worried about administration: businesses may need records identifying which portions were model-generated and which were human-authored, while the boundary between ordinary digital tools, rendering pipelines, and generative systems remains awkward in real production. Questions about lengthy prompting and iterative chatbot-based coding showed that the human-contribution test still leaves hard cases unresolved.
Where opinion split
The sharpest dispute is whether the ruling usefully blocks AI-content monopolies or merely creates an unworkable authorship audit. Supporters say withholding copyright from raw generations prevents companies from firing creators and turning cheap model output into exclusive property; skeptics say substantial human iteration, debugging, prompting, and editing can be creative work even when no one typed the final pixels or code. A related clash concerns training: critics fear courts will protect outputs while insulating companies from responsibility for copied inputs, while others stress that the cited case never decided AI authorship or broadly cleared training practices.
Coverage & discussions
Hacker News · pluralistic.net
Community discussions
Reported comments: 1 · 1 saved in sample
Too few usable comments or discussion unavailable
Community Sentiment
Positives
- Keeping raw model output outside copyright denies media companies an easy route to privatize automated content after eliminating the people who would have authored it.
- Human editing, selection, and arrangement can still earn protection, leaving room for AI-assisted production without pretending the model itself is a rights-holder.
- The thread welcomes a rule that forces companies claiming exclusive rights to identify real human creative contribution rather than treating compute spending as authorship.
Concerns
- The feared endgame is a lopsided AI economy where companies can train on creators’ work with little liability yet secure ownership over the generated products.
- Separating AI-generated fragments from human-authored changes could force expensive provenance records for every image, word, note, or code revision in commercial workflows.
- A strict focus on who produced the final expression leaves difficult cases unresolved when someone spends months directing, testing, debugging, and refining work through an LLM.
- The legal line between generative AI and ordinary computer-assisted creation still looks muddy, especially for rendering tools and heavily automated production pipelines.
