According to a new policy statement from the U.S. Copyright Office, generative AI could be eligible for copyright protection.

But, there’s a catch: Proof of human authorship (a prerequisite reaffirmed by the Copyright Office a year ago) is still required, with applications evaluated on a case-by-case basis.

If the applicant can show that generative AI was used, but then a human did something to change or significantly tweak the output, then it arguably could be judged as eligible for copyright protection.

>>Follow VentureBeat’s ongoing generative AI coverage<<

However, simply inputting a prompt into a generative AI tool, which then spits out a result, is not eligible.

Human, not generative AI, authorship is key to copyright

“The Copyright Office is saying, ‘We can’t ignore this,’ and they probably anticipate being inundated with copyright applications,” said Bradford Newman, who leads the machine learning and AI practice of global law firm Baker McKenzie in its Palo Alto office. 

The case-by-case determination of copyright issuances versus denials “will be all over the place,” he warned. “And like with patents, you’re going to need to prepare the copyright application clearly explaining what you — the human — did, vis-a-vis the generative AI.”

Still, some legal experts say it is a good first step. “This Copyright Office statement is a signaling of an open mind to further evolution,” said Jim Flynn, managing director at law firm Epstein Becker Green. That being said, the Copyright Office operates within existing laws, he emphasized — “so any breakthrough change or any fuller dispensing with the human authorship requirement would probably take legislative action.”

The policy statement, he added, is more about presenting public guidance right now. But going forward, he said he is “heartened” by the Copyright Office saying that it plans a “widespread ‘initiative to delve into a wide range’ of the copyright issues surrounding AI-generated and AI-influenced works.”

Setting the stage for complex generative AI arguments

The arguments for and against copyright protection for generative AI outputs are sure to get interesting, Newman said. For example, what about long, highly-detailed and engineered prompts that are then iterated on within a generative AI tool?

“The Copyright Office is saying ‘maybe’ as far as whether the image could be copyrighted,” he explained, “But you’re going to have to explain what you did — even with a four-paragraph prompt to create a painting, I could see them saying, you know what, what comes out is still generative art.”

These issues have already come into play, he pointed out. Just last month, Reuters reported that according to the U.S. Copyright Office, images in a graphic novel created using AI image generator Midjourney should not have been granted copyright protection.

Kristina Kashtanova, author of Zarya of the Dawn, was entitled to a copyright for the written parts of the book, as well as how the images were arranged, but not for the images themselves, the office said.

The letter said that the Copyright Office would “reissue its registration for Zarya of the Dawn to omit images that ‘are not the product of human authorship’ and therefore cannot be copyrighted.”