US feds say AI-generated prompt outputs can’t be copyrighted

If you use an AI image or text generator to make a work of “art,” does it belong to you? Or, in more relevant legal terms, can you copyright and sell the output while preventing others from selling it themselves? That’s a huge question hanging over the heads of anyone tempted to use AI tools for commercial products. And according to the latest guidance from the US Copyright Office… well, it’s complicated.

The bottom line of the updated Report on Copyright and Artificial Intelligence (PDF) is that a work of art needs “some degree of originality” and “human authorship” in order for it to be eligible for copyright in the United States. Crucially, simply plugging prompts into an AI image generator or text generator does NOT meet this burden. Because the author (or artist, or other relevant creative term) of a work is defined as “the person who translates an idea into a fixed, tangible expression,” an AI system cannot meet this burden, even though it’s using input from a human to generate its output.

“Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements,” says the report, spotted by Reuters.

According to the Copyright Office, there’s simply too much work being done autonomously between the prompt and the output to still consider it made by a person. Commenting on established case law, the report says that “…the Supreme Court has made clear that originality is required, not just time and effort.”

One of the determining factors in the report is that an AI system can create a more or less infinite amount of output, somewhat related but basically unlimited in variation, which indicates a lack of human input and control:

“The fact that identical prompts can generate multiple different outputs further indicates a lack of human control. […] The black box of the AI system is providing varying interpretations of the user’s directions. Repeatedly revising prompts does not change this analysis or provide a sufficient basis for claiming copyright in the output.”

Now, just because something created entirely by feeding prompts into an AI tool isn’t copyrightable doesn’t mean that any use of AI makes a work of art ineligible for copyright. After all, extending that argument to digital art and media tools in general—or even going back to the first uses of cameras to capture still images—would overly limit creative expression and the ability to sell it.

“The use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection for the output,” says the report. “Copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material.”

To give a recent example, Activision’s use of AI-generated still images in something as complex as the latest Call of Duty game doesn’t make the whole game ineligible for copyright. And given its association with a larger commercial product, the images themselves aren’t available for copyright-free use to other companies. Or at least that’s how I’m reading the interpretation. (I am not a lawyer, certainly not a copyright lawyer.)

So, how are we going to determine how much AI is too much? Where’s the line for when a product or project uses too much AI-generated content and thus ceases to be “human” enough for it to qualify for copyright? Frustratingly, there’s no clear line as the deciding factor. “Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis,” says the report. That feels like an echo of the famous “I know it when I see it” obscenity test, which has actual legal precedent in the United States.

The ambiguity seems inevitable to end in lawsuits. If someone makes, say, an AI tool that can create an entire mobile game from a single prompt, and someone else copies and sells that game on the basis that it can’t be copyrighted, it’ll be down to a judge to determine whether the original work has enough human creativity to qualify for protection.

A few things to remember about this situation: one, the law always lags behind technology as a matter of course; and two, this ruling only applies to the United States—and even then doesn’t have the power of law. A new law passed by Congress or an executive order from the president could change some or all of it at any time. When it comes to how AI affects copyright, we’re still very much in the weeds.