In a significant development for the evolving AI landscape, the U.S. government has established a clear benchmark regarding copyright protections for generative AI outputs, stating that without human involvement, these creations lack legal safeguarding.
As part of an extensive report on AI and copyright, the U.S. Copyright Office has concluded that unmodified outputs from generative AI tools do not qualify for federal copyright protection. This decision aligns with a comprehensive initiative to address key legal dilemmas arising amid the rapid expansion of AI technologies and their implications for copyright under the U.S. Constitution.
“Copyright protection for generative AI outputs is only applicable when a human author has incorporated sufficient expressive elements,” the report states. This includes instances where a human-created work is recognizable in the AI output or where a human artist makes creative alterations, but simply giving prompts does not suffice.
Notably, works that incorporate generative AI in their creative processes, maintaining “the centrality of human creativity,” remain eligible for copyright protection. This underscores the necessity of human contribution in artistic endeavors, as explained by the U.S. Copyright Office. “Providing legal protection for materials predominantly determined by machines would undermine the constitutional objectives of copyright,” the report emphasizes.
This ruling implies that images or videos generated by AI technologies cannot be copyrighted by the individuals who utilize them, regardless of the complexity of the prompts provided. Furthermore, prompts themselves are not eligible for copyright, nor are the various iterations of generated content.
The report also outlines guidelines regarding the requisite level of human involvement in AI-assisted art creation, particularly noting the creative processes in filmmaking. A crucial aspect of this decision is the unpredictability of generative AI outputs, which can yield varied results from identical or similar prompts.
“While entering prompts into a generative AI system might seem akin to briefing a human artist, fundamental differences exist. A human commissioner can supervise, guide, and comprehend the contributions of a human artist,” the report elaborates. “The discrepancies between prompts and outputs illustrate that users lack control over how their ideas manifest.”
A subsequent section of the report, set for release later this year, will further explore the ramifications of copyright concerning the training of AI models and tools using original, protected works.
Topics
Artificial Intelligence
