Basic Concepts

There are two basic requirements that must be met for something you've created to be eligible for copyright: it must be fixed in some tangible form and it must be original.

While there are two requirements, there are at least four ideas: create, fixed, tangible form, and originality.

You can’t create an idea or a fact, so, no copyright.

Tangible form means that your creation can be expressed to others, usually through visual or other means. Fixed means that the tangible form can’t be ephemeral, which generally means there must be a physical record. Performance art, for example, usually has a tangible form that, by definition, is not fixed and hence not subject to copyright. Same for ice sculptures.

However, it’s easy to fix a story in tangible form. Writing it on a piece of paper is sufficient. If you type your story on your computer, that puts it in tangible form. It’s fixed once you save it. It’s that easy. Just remember to save it, otherwise it’s not fixed and vanishes when you turn your computer off!

“Original” means that the “tangible form” is something you alone created. In particular, it means the tangible form is not copied in whole or part. The original creation doesn’t have to be novel (as in patent law), unique, imaginative, or inventive. Creating it doesn’t even have to be hard to do. It can even be obvious. It just means that you, on your own, set the fixed, tangible form. Under copyright law, “originality” is a low bar, and almost all works will satisfy it.

It's a low bar, but there are still restrictions. Your original work can’t copy someone else’s work in whole or part, even if what you copy is not itself copyrighted. You can't copy Homer's Iliad, even though the concept of copyright was unknown to the Greeks. If it’s not original, you can’t assert copyright.

Remember, an idea can’t be copyrighted. That means you can write a detective novel about a world-weary detective in San Franscisco, and not violate Dashiell Hammet’s copyright. However, you can’t use all or part of someone else’s tangible expression in your work, at least not without permission. Sam Spade is part of Hammet's tangible expression of The Maltese Falcon, and you can’t use him in your work without permission.

While your tangible form must be original, that doesn’t mean you can’t use tools to produce it. Most authors do research, use dictionaries, encyclopedias, and other tools. But these are tools the author deploys as part of the process of creating the tangible form.

You can follow the outline Joseph Campbell provides in The Hero With A Thousand Faces to plot your novel. That outline is an idea, and hence not subject to copyright. It’s an idea that can and has been used by many authors, which was Campbell’s whole point. Your novel is one of many possible tangible expressions of that idea, and you can own the copyright to it.

In fact, you can use any outline to plot your novel. Outlines are ideas and not subject to copyright. The tangible form, the novel, is.

You can write a sequence of stories about travelers telling stories to each other, even though Chaucer has already had that idea. The idea doesn’t have to be original, just your tangible expression of that idea.

You can write a detective story, even though Edgar Allen Poe came up with the idea.

Continuing on this line, you can use technology to assist in the creation of your tangible expression. You don’t have to use a quill pen and parchment to write your novel. You can use a typewriter, a computer, or Microsoft Word. You can use a spell checker, a grammar checker, or other tech tools to improve or refine your tangible expression. But your hand, the creator’s hand, must set the final form of the tangible expression.

Some forms of technology are fine, but what about artificial intelligence? We’ll get to that, but the US Copyright Office report asserts that there is ample existing case law to handle questions regarding AI generated content. The advent of photography and the consequent evolution of case law is an example.

The example of photography

In the nineteenth century, many artists made a living doing portraits. Such portraits had fixed tangible form and were original, and thus the artist who created them owned the copyright.

Then photographs got invented. Photographs also have tangible form. But did the photographer “create” them, or was the final form set by the camera? Today, the final form of your selfie is entirely set by your phone—all you do is click a button. How creative is that?

The answer, at least in the US, is that it’s sufficient to assert copyright. The person who clicks the button is, generally, the creator and thus the owner of the copyright. “Person” here is critical—a monkey can push the button and take a selfie, but US Copyright Office confirms that the law only applies to human-created works.

In some cases, though, someone else might be creatively involved in the photograph and own the copyright rather than the person who clicks the button. There might be decisions about staging, costuming, lighting, or even inducing facial expressions, that constitute “setting the final form” before the button gets pushed.

The point is, though, that when you click the button for your selfie, you’re generally the one making those decisions. You’re exercising the minimal creative control over the photograph required to meet the low bar of originality. Since the photograph is both fixed and tangible, it qualifies for copyright protection.

If a monkey clicks the button, however, there’s no copyright. The law only applies to works created by humans. Similarly, if a security camera takes a picture, that’s an entirely automated process with no human agency, so no copyright.

You can no doubt see where this is headed. We’re ready at last to discuss the most recent report.

Generative AI and Copyright

The first finding in the January report is that works wholly created by generative AI with no human agency do not accrue copyright protection. The monkey precedent establishes that. A big corporation running a super-duper AI program on their trillion-dollar computers might use it to produce a novel a zillion times better than War and Peace. (Probably not, but bear with me.) They can’t copyright that novel and make a fortune off of it because a computer wrote it. The novel, the tangible form, involved no human agency. It’s exactly like the monkey’s selfie. Not much motive for a profit-hungry mega-company in that.

But what about the prompt that generates AI content? Can the author of the prompt claim copyright on the prompt itself, or on the resulting content? The report notes that a prompt alone is insufficient to claim copyright on either the prompt or the result. Via 17 USC § 102(b), copyright does not protect "any idea, procedure, process, system, method of operation, concept, principle, or discovery, regardless of the form in which it is described, explained, illustrated, or embodied...'' in a work, and thus the prompt cannot be protected by copyright. It's similar to things like recipes or knitting patterns that also aren't generally protected by copyright.

A “prompt” is like an outline or an idea, and hence you can’t claim copyright on the prompt. If the tangible form resulting from that prompt is entirely AI generated, then it is not subject to copyright for the reason noted above.

This finding is also based, at least in part, on the observation that minor changes to prompts can result in unexpected and significantly different outputs. The person writing the prompt is thus exercising insufficient creative control over the tangible form. The technology is evolving, however, and the report notes that this finding may change along with the technology.

What if you use AI to generate an outline for your novel, then you write the novel? Well, at least in principle, that’s like using the outline from A Hero With a Thousand Faces. When you use it to write your novel, you’re setting the tangible form of the idea, i.e. the outline. The report finds that you can claim copyright on the resulting novel.

Where it gets tricky is when AI-generated content is embedded within your own, original content. Suppose, for example, you’re flummoxed trying to come up with a description of the Sahara Desert and ask ChatGPT to write a description for you? If you put that in your novel, does that compromise your copyright?

More generally, what, if any, is the line between spell checking your novel and having ChatGPT write content? How about if you use what ChatGPT wrote, but you revised it? Does changing one word make it original, or does the change need to be more substantial? How much AI-generated content would it take to void the copyright protection for an entire novel?

The Copyright Office’s report concludes that, at least for the moment, there is no bright line that answers these questions. Indeed, it cautions against establishing such a bright line. Copyright disputes over nuances like this are common and rely on case-by-case evaluation and the application of appropriate precedent and judgement, not rigid rules. The report finds that existing law and relevant processes will be sufficient in resolving questions regarding mixed content--content that is partly AI-generated and partly original.

An example of "relevant process" is the famous dispute between Led Zeppelin and Spirit over a chord progression in Stairway to Heaven. You can’t copyright a chord progression, but can copyright a song. The extent to which one song might or might not violate another artist’s copyright is a matter of judgement, not the mere use of a chord progression, no matter how similar the songs may sound.

The case-by-case assessment on how the use of AI impacts copyright in any document is likely to hinge on the creative process—whether the AI is used to assist in polishing the document or, instead, acts as the principal decision maker in fixing the final form of the document. Authors who plan to use AI as an assistive technology will likely need to attend, at least for the present, to documenting their creative process.

Some AI platforms permit the author to interact with the AI-generated content, editing and revising it. The report reiterates that, in this case, documenting the creative process will be critical in assessing the extent, if any, to which copyright might attach.

Name, Image, and Likeness

The report on digital replicas is new since my last newsletter, so I want to briefly discuss this topic.

What if someone takes a picture you? Do you own the copyright of your own image? The answer, in most cases, is no. However, most US states have laws protecting your name, image and likeness (NIL), and a third party can’t use those without your consent, even though they might own the copyright on a photo of you. If a third party profits from using your name, image, or likeness, you may usually make your consent to use the same contingent on compensation. If you follow NCAA football, NIL is a now-familiar concept.

A related issue is a digitally-generated replca of your name, image, or likeness. Suppose, for example, a movie studio makes a digital copy of Brad Pitt and casts that copy in a movie? Can they do that? The technology for that isn’t quite there, but could be soon. However, the technology does exist to realistically replicate someone’s voice. You might mistake a digital replica of a distinctive voice—William Shatner’s, say—on the telephone. More frightening, you might mistake the replica of a politician’s voice for the real thing. The possibilities for damage are frightening to contemplate.

While state laws may provide some protection for NIL, there are currently no national standards. The US Copyright Office’s report of July 31 last deals with such digital replicas. The report concluded that there was an urgent need for national legislation to regulate this technology and to protect the use of name, image, and likeness for all citizens, not just celebrities.

Congress has yet to take up this issue. Doubtless they’ve been busy with more urgent matters.

Summary

In this newsletter, we discussed several conclusions from the recent reports from the US Copyright Office. These include the follwing findings:

(1) The output of generative AI is not eligible for copyright protection.
(2) The prompt that resulted in the output of generative AI is not eligible for copyright protection.
(3) Documents with mixed content--partly original to the author and partly generated by AI--might be eligible for protection in whole or part by copyright.
(4) Current statutes and established procedures are adequate to resolve issues related to mixed content and copyright (i.e., no new legislation is needed).
(5) As with the introduction of other technologies, standards for mixed content will evolve through the application of precedents arising from specific cases (case law).
(6) Documention of the creative process will likely inform decisions in legal disputes involving mixed content and copyright.
(7) New legislation is urgently needed to resolve numerous critical issues related to digital replicas.


The copyright office plans a third report regarding the databases that "teach" artificial intelligence. Many of these databases include material that is protected by copyright, and their use in producing the output from generative AI engines is is likely a violation, in whole or part, of said copyright. This third report, planned for later this year, will discuss the broad question of the use of copyrighted material in the training data for generative AI engines.

References

Copyright and Artificial Intelligence, Part 1   US Copyright office report on Digital Replicas
Copyright and Artificial Intelligence, Part 2   US Copyright office report on Generative AI and Copyright
Title 17, US Code   Copyright Law of the United States