U.S. Copyright Law won’t cover solely AI-generated art

Amid a churning AI market and increasing anxiety about the role of AI in the creation of works of art, the U.S. government has set a strong benchmark: If a human isn’t involved, it’s not worth legal protections.

Part of a far-reaching report on AI and copyright, the U.S. Copyright Office has interpreted that unedited outputs of generative AI tools don’t qualify for federal copyright shelter. The decision comes as part of a years-long AI initiative that set out to answer several outstanding legal questions as the AI boom ramped up, including if the U.S. Constitution’s Copyright Clause permits protection for AI-generated material.

“The outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements,” writes the office in the report’s second installment, issued Jan. 29. “This can include situations where a human-authored work is perceptible in an AI output, or a human makes creative arrangements or modifications of the output, but not the mere provision of prompts.”

Works that use generative AI in their creative process or that include AI-generated material are eligible, as they still retain “the centrality of human creativity” rather than having AI stand-in for human creators, Shira Perlmutter, register of copyrights and director of the U.S. Copyright Office, explained. “Extending protection to material whose expressive elements are determined by a machine, however, would undermine rather than further the constitutional goals of copyright.”

That means that images or videos created by tools like Midjourney or OpenAI’s DALL-E 3 can’t be copyrighted by their individual generators, even if said person wrote a complex, entirely original prompt to generate the content. Prompts alone aren’t up for copyright either, and neither are successive iterations on such content.

The office’s report lays out further guidelines for the “level” of human involvement in creating AI art using assistive technologies, including the use of computer generated images in the filmmaking process. A key factor in the office’s decision was the unpredictability of a generative AI output, which can produce different results with the same, or similar prompts.

“Although entering prompts into a generative AI system can be seen as similar to providing instructions to an artist commissioned to create a work, there are key differences. In a human-to-human collaboration, the hiring party is able to oversee, direct, and understand the contributions of a commissioned human artist,” the report explains. “The gaps between prompts and resulting outputs demonstrate that the user lacks control over the conversion of their ideas into fixed expression.”

A third part of the report, set to release later this year, will elucidate on the role of copyright in training AI models and generative AI tools on original, protected materials.