Copyright Office AI Report Part 2
Today the Copyright Office released Part 2 of three studies on AI copyright issues. Part 1, released on July 31, 2024, dealt with digital replicas. Part 3, due out this quarter, will deal with the use of copyright works in training models. Today's report dealt with the copyrightability of works that employ AI, so-called generative AI. Essentially, the Office called for the status quo in terms of law, and its own practices. The Office adheres to its views that human authorship is required so that if there is no such authorship then there can be no registration. I agree of course.
As regards prompts, the Office writes that prompts, alone, at least currently, are insufficient to satisfy the human authorship requirement. Why? "Prompts essentially function as instructions that convey unprotectible ideas. While highly detailed prompts could contain the user’s desired expressive elements, at present they do not control how the AI system processes them in generating the output." Page 18. The Office acknowledged that there is a craft to drafting prompts, sometimes called "prompt engineering." The Office also acknowledged that many AI platforms offer tools for iterative selecting, editing adapting output, page 25, and that prompts can be revised hundreds of times as the human creator refines his or her judgment about what the final version of the output should be. Page 17.
Yet, the Office concluded that "Prompts essentially function as instructions that convey unprotectible ideas." Page 19, while agreeing that "While highly detailed prompts could contain the user’s desired expressive elements, at present they do not control how the AI system processes them in generating the output. " Yes prompts are instructions, and even if those instructions are deemed to be ideas ("beautiful sunset in Bolinas, with whales breaching"), the real issue is whether by using human created prompts revised hundreds of times to reflect the desired output, that output is "authored" by the person who crafted the prompts in order to achieve a subjective, very human final product. I don't see how this can be doubted. Focusing on the tool to deny that the tool was used to create an original expressive output makes no sense. One is not seeking protection for the tool, but the output. To say a machine is controlling how the tool operates doesn't make sense either since the human still decides on the content of the output. In the pre-AI world, it didn't matter if when I push a button to create special effects, I have no idea how the machinery works. All that matters is if the output contains the content I want. To characterize tthe use of prompts as spinning a roulette wheel or walking into a store and saying "I like this" somehow makes one an author, see Pages 17-18, is simply not in the leats bit apposite to a process of creating prompts, editing them hundreds of times to finally get the desired result.
It is no different than a sculptor who works in large, heavy pieces of marble instructing fabricators how to make his or vision and then having cranes put everything into place. it has never been a requirement in copyright law that the author physically create the work. The requirement is that the work -- the intangible expression be the result of that person's artistic vision. I do not believe that the Office's blanket, fact-free exclusion of works created by prompts even with no post-production alteration, meshes with fundamental principles of copyright law. Yes, there may well be individual works that do not meet the requisite human involvement, and for those, as in the Thaler case, the Office was correct in rejecting the application. But when we speak of across-the-board rules, I depart from the Office's position.
Helpfully, the Office cautioned that: "There may come a time when prompts can sufficiently control expressive elements in AI-generated outputs to reflect human authorship. If further advances in technology provide users with increased control over those expressive elements, a different conclusion may be called for." Page 21. My disagreement with the Office is that we are at this time, whether some individuals rise to the requisite level or not should not affect those who do.
As for the use of AI "expressive inputs" in assisting the creation of a work that does contain the requisite human authorship, the Office continues to allow registration for those elements, if separable from non-protectible elements. It gives a helpful illustration on page 23. Quite unhelpful, though, in my opinion is this passage from page 24:
As illustrated in this example, where a human inputs their own copyrightable work and that work is perceptible in the output, they will be the author of at least that portion of the output. Their own creative expression will be protected by copyright, with a scope analogous to that in a derivative work. Just as derivative work protection is limited to the material added by the later author, copyright in this type of AI-generated output would cover the perceptible human expression. It may also cover the selection, coordination, and arrangement of the human-authored and AI-generated material, even though it would not extend to the AI generated elements standing alone.
This is disturbing on a number of levels. The work on page 23 was based on the applicant's hand-drawn illustration which was then used as an input, along with this prompt:
“a young cyborg woman
(((roses))) flowers coming
out of her head,
photorealism, cinematic
lighting, hyper realism, 8k,
hyper detailed.”
Here is the hand drawn input on the left, and the final work on the right after the prompts were used:
The report states that:
The drawing itself is a copyrightable work, and its expressive elements are clearly perceptible in the output, including the outline of the mask, the position of the nose, mouth, and cheek bones relative to the shape of the mask, the arrangement of the stems and rosebuds, and the shape and placement of the four leaves.
The final result, after the prompts were processed, was a work where the hand drawn elements were not altered; the prompts generated a colored background, texture, and three colored roses. One should not consider the resulting work a derivative work (there was no pre-existing work) or a compilation. The resulting work is an integrated work of visual art. The Office's efforts to treat such integrated works as derivative works or compilations merely because a tool was used for a portion of the work is shocking.
Making this worse, IMHO, the Office annotated the registration as follows:
“Registration limited to unaltered human pictorial authorship that is clearly perceptible in the deposit and separable from the non-human expression that is excluded from the claim."
This too denies the very nature of the work as a single artistic vision. It also raises what I consider a critical issue, the use of disclaimers for AI assisted works that are registrable. Despite strong comments by the Motion Picture Association and others objecting to this use of disclaimers, the report never discusses an issue that goes to the heart of the Office's role in the registration process. I do not believe the Office has the authority to require registrants to disclaim material they have not affirmatively claimed. If an applicant claims "idea, method, or operation," that has to be disclaimed. But if an applicant claims "entire work," the Office should not, and I believe cannot, require a disclaimer. As applied to computer programs that generate typeface, the Office abandoned that practice in 1992. I wrote the Office's abandonment of that practice as a Policy Planning Advisor to the Register and am disappointed to see it resurrected. It should not be resurrected. It is bad policy, created needless work on applicants' part, and may lead to protracted fights in litigation.
Section 602.4(B) of the Office's Compendium, describing the scope of the Office's examination, notes that the Office examines for whether the work contains copyrightable subject matter. The Office is an office of record and "accepts the facts stated in the registration materials, unless they are contradicted by information provided elsewhere in the registration materials or in the Office’s records." Compendium Section 602.4(C), Thus, where an applicant describes the nature of authorship as "entire work," does not affirmatively claim any unprotectible material and there are no statements like "the author is the Creativity Machine" and the work was " “[c]reated autonomously by machine," as in Thaler, so long as the deposit copy contains sufficient original material the Office's duty is to issue the registration and leave it up to the adversarial process to sort things out.
There is much to appreciate in the Office's report, and some that must be questioned.
Comments
Post a Comment