Copyright office recently ruled that outputs are generally not copyrightable without significant human changes. Prompts are also not copyrightable. I'm actually surprised that is not discussed around these parts more because it essentially means any app etc built by AI has no copyright protection. Thus code from llms cannot be opensourced using opensource licenses. Business cannot stop users from stealing their apps etc.
It also means that if you build something with a popular llm then big tech now has your inputs and outputs, and you have no IP to stop them from stealing either.
The next ruling from the copyright office is around fair use that big tech is pushing for.
Not looking forward to Plagiarism As A Service.
What's the status right now for whether LLM outputs can be copyrighted by the same big companies that ignored copyright to train them?
Copyright office recently ruled that outputs are generally not copyrightable without significant human changes. Prompts are also not copyrightable. I'm actually surprised that is not discussed around these parts more because it essentially means any app etc built by AI has no copyright protection. Thus code from llms cannot be opensourced using opensource licenses. Business cannot stop users from stealing their apps etc.
It also means that if you build something with a popular llm then big tech now has your inputs and outputs, and you have no IP to stop them from stealing either.
The next ruling from the copyright office is around fair use that big tech is pushing for.