Copyright Office Report Rejects New Protections for AI-Generated Content
by
February 5, 2025
Generative AI models have pushed the boundaries of existing Intellectual Property (IP) laws. AI IP issues generally fall into two camps, the IP rights attached to training data and the IP rights of AI-generated content. We’ve covered several developments related to the former, but the latter is equally important. The U.S. Copyright Office recently published guidance on this point, further elaborating on when AI outputs are subject to copyright protection. Ballard Spahr summarizes the guidance in a new memo:
- “The Office concludes that no new laws are needed to resolve questions of copyrightability and artificial intelligence.
- Copyright protects the original expression in a work created by a human author, but it does not extend to works or portions of works where a human did not sufficiently control the expressive elements.
- Creation of text ‘prompts’ to an AI system is insufficient to make a human the author of outputs.
- Humans can still claim authorship over works where AI was used as a tool, or where a human makes expressive modifications, or arranges and selects AI-generated works.”
Human input is the key element that determines whether AI-generated content is protected under copyright law. Results generated from text prompts are unlikely to meet the appropriate threshold. This holds true even when text prompts are repeatedly revised and tailored to generate a specific outcome. AI content is more likely to be protected by copyright if AI is used to edit human creations rather than create entirely new works. For example, if I ask Chat GPT to correct grammatical errors in a short story I wrote, then the output will likely be protected by copyright. However, a short story Chat GPT writes from scratch is likely not eligible for protection.