The copyright problem with AI-generated code

Why open source licenses may not apply to what you didn't actually write.

Since the advent of “Generative AI” and Large Language Models like ChatGPT, Gemini, Copilot, and Claude, many people have extolled the virtues of vibe coding. The promise of vibe coding is that you don’t have to be a developer to create software projects. Instead, you can simply describe the project in a prompt, and the AI system will write the code for you. In a 2025 keynote, Nvidia CEO Jensen Huang described vibe coding as a great equalizer.

However cool that prospect might be to some, there are some pretty big implications for open source programmers who lean too heavily on AI coding assistants. You’ve probably heard many times how open source projects that use AI to write code, or that accept contributions authored by AI, tend to “turn off” other developers. Many open source developers get “into” open source projects because they like writing code. Using an AI to write code for you is like going to the bank to watch them count other people’s money. It’s just not fun.

But instead of revisiting that issue, I’d like to address the legal implications of using AI to write code.

Why open source licenses can’t exist without copyright

All open source licenses, going back to the original “Free Software” license, the GNU General Public License, rely on copyright to establish their provisions. GNU GPL version 1 refers to copyright several times, including in the preamble and body. Here are a few key instances:

We protect your rights with two steps: 
 (1) copyright the software, and 
 (2) offer you this license which gives you legal permission 
     to copy, distribute and/or modify the software.

...

This License Agreement applies to any program or other work 
which contains a notice placed by the copyright holder saying 
it may be distributed under the terms of this General Public 
License. The "Program", below, refers to any such program or 
work, and a "work based on the Program" means either the 
Program or any work containing the Program or a portion of it, 
either verbatim or with modifications. Each licensee is 
addressed as "you".

You may copy and distribute verbatim copies of the Program's 
source code as you receive it, in any medium, provided that 
you conspicuously and appropriately publish on each copy an 
appropriate copyright notice and disclaimer of warranty; 
keep intact all the notices that refer to this General Public 
License and to the absence of any warranty; and give any other 
recipients of the Program a copy of this General Public License 
along with the Program. You may charge a fee for the physical 
act of transferring a copy.

Copyright is the foundation of an open source license. It gives you, the copyright holder, the right to define how others can copy your work. If copyright did not exist, you would not be able to place your software under an open source license, such as the GNU GPL or any other.

Read more: You don’t know what you’re actually shipping

AI-generated content and the copyright problem

US courts are increasingly finding that AI-generated content is not eligible for copyright protection, which raises issues for open source licenses. In one well-known case, an artist used AI to generate an image, Théâtre D’opéra Spatial, and entered it into an art contest without disclosing that it was AI-generated.

The piece won the contest, but the artist was ultimately unable to secure copyright for it. As highlighted in a 2023 public radio article, the artist applied for copyright protection and was turned down by the US Copyright Office:

The image was striking enough to win part of the Colorado State Fair's 
art contest last year. But the U.S. Copyright Office on Tuesday issued 
a final decision that it can not be copyrighted, meaning Allen can't 
get legal protections to stop others from reproducing and using the 
image themselves.

The federal office found that Allen is not the "author" of the piece. 
Instead, the office wrote, "his sole contribution to the Midjourney 
Image was inputting the text prompt that produced it," and then making 
visual modifications afterward. The office has found previously that 
copyright law is meant to protect human authors, not machines.

In earlier guidance, the office wrote: "[W]hen an AI technology 
receives solely a prompt from a human and produces complex written, 
visual, or musical works in response, the 'traditional elements of 
authorship' are determined and executed by the technology—not the 
human user."

The Copyright Office pointed to numerous other cases in which the 
courts have rejected non-human authors, including one involving a 
book purportedly authored by "non-human spiritual beings." (The 
decision also referenced a dispute about whether a monkey's selfie 
could be copyrighted.)

I encourage you to read the US Copyright Office report on Copyright and Artificial Intelligence, specifically Part 2 on “Copyrightability.” At 52 pages, this might seem like a long document, but many pages are half-filled with footnotes, so the effective page count is closer to 25 or 30.

Page 8 of the report mentions the Théâtre D’opéra Spatial piece and the court decision that addressed claiming copyright on AI-generated work (bracketed numbers refer to footnotes in the report):

In 2023, the U.S. District Court for the District of Columbia became 
the first court to specifically address the copyrightability of 
AI-generated outputs. [37] The plaintiff challenged the Office's refusal
 to register an image that was described in his application as 
"autonomously created by a computer algorithm running on a machine." [38] 

Affirming the Office's refusal, the court stated that "copyright law 
protects only works of human creation," and that "human authorship is 
a bedrock requirement of copyright." [39] It found that "copyright has 
never stretched so far [as] ... to protect works generated by new forms 
of technology operating absent any guiding human hand." [40] Because, 
by his own representation, the "plaintiff played no role in using the 
AI to generate the work," the court held that it did not meet the human 
authorship requirement. [41] The decision has been appealed. [42]

The key takeaway from the report is that there’s no hard-and-fast rule for how much AI a person can use and still claim copyright. These issues sometimes need to be determined on a case-by-case basis. The general guideline: the more decisions a human makes in the process, such as arranging the AI-generated output, the more likely that human can claim copyright. For example, the report says this on pages 10-11:

Although an AI-generated output cannot be considered a joint work 
with respect to the user and AI system, [57] joint authorship provides
 a helpful analogy in assessing whether a party contributed sufficient 
expression to be considered an author. [58] To be a joint author, one 
must make a copyrightable contribution. [59] "A person who merely 
describes to an author what the commissioned work should do or look 
like is not a joint author for purposes of the Copyright Act." [60]

In other words, just because you described it doesn’t mean you have any claim to the copyright. Think of it this way: if you hire an artist to paint your portrait and make specific requests like “make my nose look smaller” or “make the background look like a garden,” that doesn’t make you a “co-painter” of the piece. You simply described the end product; someone else actually did the work.

Where that leaves your vibe-coded project

That’s the dividing line in AI-generated work, and it’s why open source developers should be mindful of how much they rely on AI in their projects. The more you lean on AI to write code for you, the less likely you are to hold copyright on the result. In the extreme case of “vibe coding,” where you only describe the end product and the AI does all the work, you may not be able to claim any copyright at all.

And because open source licenses rely on copyright, you may not be able to place a “vibe coded” project under an open source license, such as the GNU GPL.

More from We Love Open Source

• 15 open source backup solutions to protect your data
• 4 practical pandoc techniques for everyday document conversion
• You don’t know what you’re actually shipping
• 5 forces driving DevOps and AI in 2026
• Who benefits most from AI right now?