I happen to have two unrelated friends who work for the Library of Congress, and both of them came to my house last night. I spoke with one of them briefly about copyright law and LLM output -- he brought up the topic as one of the frustrations he's having at the office, so I followed up with a few questions for clarification.
His frustration -- people want to register copyrights for works that they prompted an LLM (such as ChatGPT) to produce. But currently copyright law does not recognize LLM outputs as protected works. To have a copyright, the output must be the work of a human.
I hadn't thought about this question at all as I was collaborating with ChatGPT and Nomi.ai to produce my 2024 NaNoWriMo story. I don't care about copyrighting my writing because I'm not trying to sell it and I'm not aware of anybody else trying to sell it as their own work.
I'm curious about editing LLM output, however. If I edit an LLM's output, then does it become my own work? How much editing would be required to transform something from not copyrightable, to copyrightable?
Practical Takeaways for Your NaNoWriMo Project
- If you're not concerned about copyrighting your story, you don't need to worry about this issue. Your focus can remain on creating something you're proud of, regardless of legal protections.
- If you want to assert copyright over your work, make sure your edits and contributions demonstrate clear human authorship. For example:
- Use AI output as raw material and rewrite it significantly.
- Combine AI-generated sections with your original ideas and writing.
- Create an overarching narrative, structure, or theme that is uniquely yours.