Summary of https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf

This report from the U.S. Copyright Office examines the intersection of copyright law and artificial intelligence (AI), specifically focusing on the copyrightability of AI-generated works. The report analyzes different levels of human involvement in AI-generated content, considering factors such as prompts, expressive inputs, and modifications.

It concludes that existing copyright law is sufficient to address these issues, emphasizing the crucial role of human authorship.

• Copyright law does not protect AI-generated works, unless there's enough human input. This isn't just about effort, but about creative input.
• The "black box" nature of AI systems is a core issue. Even developers often don't know how AI models generate their outputs, making it difficult to claim human authorship over those outputs.
• Prompts, even detailed ones, usually don't provide enough control for copyright because the AI interprets and executes them in unpredictable ways. The system fills in the gaps, and the user's control is indirect. It is difficult to demonstrate sufficient closeness between "conception and execution".
• Iterative prompting (revising prompts and re-submitting) does not equate to copyrightable authorship. It is like "re-rolling the dice" and does not demonstrate control over the process.
• "Authorship by adoption," where someone claims ownership of an AI output just because they chose it, is generally not recognized. The act of selecting an AI-generated output from many options is not considered a creative act.
• Expressive inputs, like a user's own artwork used as a starting point for AI generation, can be protected. The copyright would cover the human expression that is perceptible in the final output.
• Modifying AI-generated content can create copyrightable material. This includes creative selection and arrangement, or making sufficient changes to the AI output.
• AI is a tool, and using it does not negate copyright, if there is sufficient human creative contribution.
• There is a concern that an increase in AI-generated outputs will undermine the incentive for humans to create.
• Many countries agree that copyright requires human authorship. But, there is ongoing discussion regarding how to apply this to AI-generated works.
• There is debate on whether a sui generis right is necessary. Most commenters opposed it, noting that AI systems do not need incentives to create.
• The Copyright Office is monitoring technological and legal developments to determine if conclusions need revisiting.

The report also explores international approaches to AI and copyright, noting a general consensus on the need for human authorship. Finally, it evaluates policy arguments for legal changes, ultimately recommending against legislative alterations.