New AI Memo Just Dropped

They say: Scratch a lawyer and you’ll find a nerd.

WELL GET PUMPED NERDS because the Copyright Office just released a statement of policy on the subject of AI-generated works! I have attached the original memo to the bottom of this post, in case you prefer PDF/letter format to an unstoppable wall of text.

For those of you who, like me, have been wading through the emerging caselaw and piecing the puzzle together for yourself, this won’t have any surprises. But for anyone who wants a clear and direct statement o f current Copyright Office policy, that is exactly what this is. Here are the big points, followed by my personal digressions.

Human Authors Required

In the same vein as the recent decision to deny copyright on the individual AI-generated images in a comic book, the Copyright Office says in very clear terms that a human being is required to be an author. Crucially, they ground this first in Article 1, Section 8, Clause 8 of the U.S. Constitution (the Patent and Copyright Clause), and state that an “author” has to be a human being.

By taking a hard-line anthropocentric approach to this issue, the Copyright Office forestalls any questions about potential future AIs and their creativity. Even if an AI were created that equaled or exceeded a human being’s creative abilities, it would not count as an author because it would not be a human being.

To support that position, the memo cites some caselaw that I hadn’t seen before. In addition to the Monkey Selfie case that usually gets cited, they cited a Ninth Circuit decision in Urantia Foundation v. Maaherra, denying copyright for a book that was ostensibly written by divine inspiration. The Court had a few great quotes, but this is the core of it:

We agree with Maaherra, however, that it is not creations of divine beings that the copyright laws were intended to protect, and that in this case some element of human creativity must have occurred in order for the Book to be copyrightable.

114 F.3d 955, 958 (9th Cir., 1997)

In the Monkey Selfie example and others like it, it was possible to say that the monkey wasn’t able to exercise the required creativity necessary for authorship. That logic leaves open the possibility that some future AI with human-level creativity could hit the mark. But in the Urantia case, parts of the text were assumed to come from “the Divine Counselor, the Chief of the Corps of Superuniverse Personalities, and the Chief of the Archangels of Nebadon.” The Ninth Circuit never questioned whether those supposed being had the requisite level of creativity, and instead just relied on the fact that they weren’t human. See id.

The Ninth Circuit obviously isn’t the final say in these things, and this is just an example I picked because I thought it was funny, but I think it’s pretty instructive for how more advanced AIs will be treated, absent some basic change to the law.

The Client/Artist Analogy

The memo makes a pet theory of mine into official policy: Giving a prompt to an AI is akin to a client commissioning an artist:

Based on the Office’s understanding of the generative AI technologies currently available, users do not exercise ultimate creative control over how such systems interpret prompts and generate material. Instead, these prompts function more like instructions to a commissioned artist—they identify what the prompter wishes to have depicted, but the machine determines how those instructions are implemented in its output.

Memo, p. 7

And just like in the client scenario, the client’s bare prompt does not give the client any rights in the artist’s output. Since the output here is generated by a machine, the Copyright Office concludes that the output is therefore not a work of human authorship, and that the material is not protected by copyright.

Human Edits to AI Outputs

The memo lays out a couple of ways that AI-generated works can be shoe-horned into a copyright registration. The first is to provide a human-generated arrangement of the works into come collected form, similar to what was addressed in the comic book decision. The second is that a human may modify the original work in such a way as to introduce the requisite level of human creativity, which makes the modified work protectable. In doing this, the Office distinguishes from the use of digital tools to create and edit works, noting that the use of such tools is recognized as being part of the creative process. See Memo, p. 8.

New Requirement for Authors

The Memo finishes up by providing guidance to authors who want copyright protection over works that used AI. In particular, they instruct applicants to identify and disclaim the contributions that the AI made. See Memo, p. 9. This is entirely self-reported on the part of the author, so it opens the possibility for fraud.

The Memo therefore finishes up by explaining what dire consequences await if you try to lie about whether an AI helped make a given work. Those consequences basically just include cancellation of the mark if the Copyright Office gets wind of it or, more likely, by the Federal Courts if you ever try to sue someone on it. I expect copyright lawsuits will take on a whole new dimension as defendants try to prove that the plaintiff used an AI and didn’t disclose it.

Conclusion

So that’s your up-to-the-minute breaking AI law news! Humans remain supreme over artlaw for the foreseeable future.

One final touch: I want to emphasize that this discussion has absolutely nothing to do with the quality of the product and everything to do with the human creativity guiding the expression of an idea, no matter how meager that creativity might be. The images above were generated by AI and have no copyright protection, despite being pretty awesome.

In contrast, I just made this rendition of the “excitable young lawyer with glasses holding a legal memo” using MS Paint, so I own the copyright on it, and I absolutely forbid you from copying it. Don’t do it!

In other words, AIs are good, but they definitely have not yet reached human-levels of obnoxiousness.

Oh, and here’s the original memo: