The U.S. Copyright Office's just released a highly interesting report, "Copyright and Artificial Intelligence, Part 2: Copyrightability,". The report, drawing on a wide range of public comments, legal precedents, and international perspectives, concludes that current copyright law is sufficient to address the copyrightability of AI-generated works, and legislative change is not necessary at this time.
I. Foundational Copyright Principles and AI
- Human Authorship as a Cornerstone: The report reaffirms the fundamental principle that copyright protection in the U.S. requires human authorship, stemming from the Copyright Clause of the U.S. Constitution. This principle is not merely a formality but rather a reflection of the purpose of copyright law, which is to incentivize human creativity.
- Expression vs. Idea Dichotomy: The report emphasizes that copyright protects the expression of ideas, not the ideas themselves. This distinction is critical when analyzing AI-generated outputs, where prompts or other inputs might represent an idea, but the output is the result of the AI's expression of that idea.
- Originality Standard: The report underscores the need for originality, requiring that a work be independently created by a human author with a minimal degree of creativity. "Sweat of the brow" alone is not sufficient to establish copyright.
II. Human Contribution to AI-Generated Outputs
- AI as a Tool vs. Autonomous Creation: The report draws a key distinction between using AI as a tool to assist human authors and AI autonomously generating content. When AI serves as a tool, the resulting works can be copyrightable provided the human author's contributions meet the authorship criteria.
- Prompts as Unprotectable Ideas: The report concludes that prompts alone, even if detailed, do not typically confer copyright in the resulting AI-generated output. Prompts are considered instructions that convey ideas, and do not provide the requisite control over how the AI expresses those ideas. The "black box" nature of AI systems means that a user's mental conception is not directly translated to the output. The report also notes that, in some instances, the AI system modifies or rewrites the prompts. The report cites the lack of predictability in outputs from the same prompt as further evidence that users lack the necessary control for authorship. Iterative prompting is not sufficient for authorship, as the user is still selecting from outputs they do not control.
- Expressive Inputs as Copyrightable Contributions: Human-authored expressive inputs, such as an original drawing or audio recording that are used as inputs to an AI system, can be protected to the extent they are perceptible in the AI-generated output. This protection is analogous to that of a derivative work, extending only to the human-authored elements.
- Creative Modification and Arrangement of AI Outputs: The report acknowledges that a human author can obtain copyright in the creative selection, coordination, or arrangement of AI-generated material, as well as creative modifications of those outputs. This is consistent with compilation and derivative work principles, extending to the human contributions, but not the underlying AI-generated material.
III. International Perspectives
- Global Consensus on Human Authorship: The report notes that several countries, such as Korea, Japan, China, and EU member states, have expressed similar views on the need for human authorship in copyright law.
- Variations in Approach: However, the report observes that the specific application of authorship principles to AI-generated content varies globally.
- Monitoring International Developments: The Copyright Office is closely monitoring international developments to determine whether these approaches may ultimately converge or diverge from the U.S. stance.
IV. Policy Considerations and Arguments for Legal Change
- Incentives for Human Authors: The report argues that copyright is intended to incentivize human creativity, not the autonomous output of AI systems. The Copyright Clause is meant to promote progress by securing rights for authors, who are human creators. The report argues that AI systems do not require copyright incentives. The report expresses concern that providing copyright protection for purely AI-generated works could undermine incentives for human creation, potentially flooding the market with AI content and diluting the value of human expression.
- Sui Generis Rights: The report finds that the case has not been made for new sui generis rights to protect AI-generated content and cites the potential for this to undermine existing copyright law.
- Empowering Creators with Disabilities: While acknowledging that AI can be a valuable tool for creators with disabilities, the report emphasizes that such uses are considered assistive, and copyright protection would lie in the human author’s expression.
- International Competition Concerns: The report notes concerns about international competition and countries with more AI-friendly policies but asserts that the U.S. should not abandon its own copyright principles because other countries may not share them.
- Need for Clarity: The report acknowledges the need for clarity regarding copyright and AI, but does not believe that legislative change is necessary at this time. The report itself seeks to provide guidance to clarify the application of copyright law in this area, noting that the courts will also provide future guidance.
- Existing Law Sufficient: The Copyright Office maintains that existing copyright law, as currently interpreted, is sufficient to address the copyrightability of AI-generated outputs.
- Case-by-Case Determination: The report emphasizes the need for a case-by-case analysis of human contribution, underscoring that there is no bright-line rule for AI-generated content.
- Ongoing Monitoring: The Copyright Office will continue to monitor technological and legal developments and provide guidance. This includes updating the Compendium of U.S. Copyright Office Practices.
In conclusion, while AI is a potent tool, copyright protection remains fundamentally tied to human creativity and control over expressive elements, not just the generation of outputs. The report stresses the importance of understanding how the law applies in the specific facts of each case, and that the level of human control over expressive elements is the key factor.
