.PNG){kind=logo}
Review Board clarifies human-content requirement in SURYAST
On December 12, 2023, the Review Board of the U.S. Copyright Office refused registration of a two-dimensional artwork titled “SURYAST” in one of the first applications of its AI guidance. The registrant, Mr. Sahni, identified himself and “RAGHAV Artificial Intelligence Painting App” as authors of the work, explaining that Sahni had input a photo he had taken to the AI, and prompted it transform the photo in the style of Vincent Van Gogh’s painting, The Starry Night. Finding that the new work was generated by the AI, the Board decided “the derivative authorship [wa]s not the result of human creativity or authorship and therefore not registrable.” Ruling at 3. (internal citations omitted).
Sahni argued that by selecting the image inputs and the degree to which the filter strength of the Van Gogh style should be applied to his photo, he had sufficiently contributed to the work generated by the AI to be an author. The Review Board disagreed, finding that instructing the AI to apply a “Van Gogh” filter, and selecting the degree of filtering, Mr. Sahni’s admitted contribution to the final work, amounted to contributing the idea of a Van Gogh style version of his photo, rather than a contribution to the expression itself. See Mazer v. Stein, 347 U.S. 201, 218 (1954) (“Holder of a copyright to a lamp embedding statuettes of human figures may prevent use of the same figures but may not prevent lamp makers from incorporating different figures into lamps because the copyright does not extend to the idea of a lamp incorporating statuettes”).
The ruling in Mr. Sahni’s case was consistent with the Board’s previous rejection of the registration of a work titled Théâtre D’opéra Spatial in which Mr. Allen, the registrant, provided over 600 text prompts to an image generation AI. The Review Board observed in that case that such prompts may be creative, but that to qualify as an author in a visual work, the author’s contribution muse actually form the generated images. Allen’s refusal to limit his claim to exclude protection of the non-human authorship elements of the image was also a deciding factor.
Under the current state of the law, when considering the validity of a copyright registration in a work that includes AI-generated content, the analysis focuses on the following:
- Was the contribution of the AI more than de minimis or appreciable. To answer this question, ask whether the output of the AI, standing on its own, be copyrightable if created by a human author. In other words, would it meet the originality threshold set by the Feist case. If so, the contribution is appreciable. Alterations such as sharpening and balancing colors, tint, or tone, blurring identifying features, or replicating an edit across frames of an image may be viewed as mechanical and treated as de minimis contributions.
- Did the human author(s) make contributions to the work that were more than de minimis?
If both the AI and the human contributions are appreciable, then the Copyright Office has given specific guidance as to how to identify and disclaim the contribution of the AI in a copyright registration.
Because a copyright cannot be enforced in the U.S. without a valid registration, copyright owners should take care to ensure they have followed the guidance of the copyright office before bringing an infringement suit. The Copyright Office has explained that the disclosure requirement will be met if the AI contribution is identified in a relatively general manner, and the guidance also spells out how to correct an earlier registration to meet the disclosure and disclaimer requirements.