As much as I wish this was true, I don’t really think it is.
If the AI generated code is recognisably close to the code the AI has been trained with, the copyright belongs to the creator of that code.
I may be wrong but I think current legal understanding doesn’t support this
Under U.S. law, to prove that an AI output infringes a copyright, a plaintiff must show the copyrighted work was “actually copied”, meaning that the AI generates output which is “substantially similar” to their work, and that the AI had access to their work.[4]
I’ve found a similar formulation in a official German document before posting my above comment. Essentially, it doesn’t matter if you’ve
“stolen”copied somebody else’s code yourself and used it in your work or did so by using an AI.
The US copyright office confirms this. https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf
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. Copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material.
I’m not sure where you get that from, I’m pretty sure vibe coding still complies with these indications
“AI-generated” works can be copyrighted. However, on the condition that the AI-generated elements are explicitly mentioned in the “Excluded Material” field. In other words, the parts generated by AI are not protected, only the parts that are expressed by human creativity. Courts in the U.S have already rejected registration for many AI works because of that. Regardless, it’s still a contentious matter.
P.S. I am completely opposed to (generative) AI as well as the copyright system. I’m just stating my findings researching the law and court cases.
As mentioned elsewhere in this thread it won’t matter either way unless tested in court and that will never happen for most companies.
Did you even read your own report? It says that AI works are copyrightable in certain circumstances, not that they make a whole project public:
Copyright law has long adapted to new technology and can enable case-by- case determinations as to whether AI-generated outputs reflect sufficient human contribution to warrant copyright protection. As described above, in many circumstances these outputs will be copyrightable in whole or in part—where AI is used as a tool, and where a human has been able to determine the expressive elements they contain. Prompts alone, however, at this stage are unlikely to satisfy those requirements.
“AI-generated” works can be copyrighted. However, on the condition that the AI-generated elements are explicitly mentioned in the “Excluded Material” field. In other words, the parts generated by AI are not protected, only the parts that are expressed by human creativity. Courts in the U.S have already rejected registration for many AI works because of that.
P.S. I am completely opposed to (generative) AI as well as the copyright system. I’m just stating my findings researching the law and court cases.
The part that is untrue is the “public domain” part. If you generate code then you don’t own it because the actual human work that went into creating it was done by the owner of the AI Model and whatever they trained on.
That sounds like complete bullshit to me. Even if the logic is sound, which I seriously doubt, if you use someone’s code and you claim their license isn’t valid because some part of the codebase is AI generated, I’m pretty sure you’ll have to prove that. Good luck.
If there was an actual civil suit you’d probably be able to subpoena people for that information, and the standard is only more likely than not. I have no idea if the general idea is bullshit, though.
IANAL
You forgot the heart
I ♥️ ANAL
Would that be North African Lawyer, or North American Lawyer?
In any case, we’re splitting the cheque. /s
Are you suggesting that lawyers migrate?
Hmm, it would require a pro hac vice admission.
There was a case in which a monkey took a picture and the owner of the camera wanted to publish the photo. Peta sued and lost because an animal can’t hold any copyright as an human author is required for copyright.
https://en.wikipedia.org/wiki/Monkey_selfie_copyright_dispute
As you also find in the wikipedia article, this case is used to argue that ai generated content is not by an human author and consequently not copyrightable.
I’d argue that this is a different scenario, as AI is a tool, not a being. At least at this point.
A complex tool, but really just a tool. Without the human input, it can’t do shit.
There’s already rulings on this holding that the prompt for all LLM or image generator isn’t enough to count the result as the human’s expression, thus no copyright (both in USA and other places)
You need both human expression and creative height to get copyright protection
I’d argue that it is wildly different to vide coding.
Exactly. If I use online Photoshop or whatever, and I use the red eye removal tool, I have copyright on that picture. Same if I create a picture from scratch. Just because someone like OpenAI hosts a more complex generator doesn’t mean a whole new class of rules applies.
Whomever uses a tool, regardless of the complexity, is both responsible and benificiary of the result.
Not quite how copyright law works. Photoshop and similar gives you copyright because it captures your expression.
An LLM is more like work-for-hire but unlike a human artist it doesn’t qualify for copyright protection and therefore neither does you
Well, not how USA copyright works, but point well taken. It seems I was too naïve in my understanding of copyright.
Anything built by AI/LLMs should be FOSS by law. Oh I dream of the day.
Your wish is granted.
But you can only view the source code through an LLM
a finger on the monkey paw curls
What a vibe
Hitachi Vibe!
It already is, vibe coders cannot hold copyright on AI-generated code
Not to be pedantic, not holding copyright ≠ FOSS.
FOSS explicitly means that the developer has a copyright and is explicitly giving a license for people to use it with FOSS provisions.
It would be more accurate to say AI Vibe code is in the public domain.
Public domain code is a subset of FOSS code, so I don’t think that was inaccurate
In general terms, it isn’t really. Or at least it is a controversial topic still subject to discussion.
https://opensource.org/blog/public-domain-is-not-open-source
As to this specific topic, the fact that all of the code has to be open source is part of the 10 criteria https://en.wikipedia.org/wiki/The_Open_Source_Definition
As such, you can’t consider open source the public domain portions of a codebase that also has proprietary portions.
That’s not entirely true, it doesn’t make it FOSS.
-
Vibe coders sign a contract when they use AI to generate stuff, which gives rights away to the company. Regardless of copyright protections from the state, a contract is in most cases legally binding.
-
Copyright law requires human authorship as opposed to random generation. This doesn’t inherently exclude all generative works, algorithms that were carefully crafted and datasets curated can potentially have their results considered “authored” but the AI Company owners that made them.
In order to make it true we need to pass laws that regulate the AI companies and their slop. In the meantime, I recommend nobody uses slop code. Actually, I’d recommend that regardless of ownership rights.
I signed no contracts to download the open source local models I use for code generation, just for what it’s worth
Ah yes, that model. Of course. You’re right. I would know because I have read the details for every model in existence and can clearly infer which one you’re talking about. Yes.
There’s lots of open source models you can download from Hugging Face, Ollama, and even github without signing any contracts or terms of use. Gemma3, Llama, Ministral, GLM, olmo, and a bajillion others. GLM-4.7-Flash is a very capable agentic model that can run at very usable speeds on commodity hardware - and none of what it generates is dictated by any agreements or policies agreed to anywhere.
https://ai.google.dev/gemma/terms
You agreed to these terms when you downloaded Gemma3 from an official source and if you did not then you’re an unlicensed user AKA piracy.
Interestingly, none of the official sources for the model weights clickwrap the download in a way that forces the user to read or agree to those terms before downloading. There is precedent for such terms being unenforceable when the user isn’t forced to agree to the terms.
-
Aren’t you all forgetting the core meaning of open source? The source code is not openly accessible, thus it can’t be FOSS or even OSS
This just means microslop can’t enforce their licenses, making it legal to pirate that shit
It’s just the code that’s not under copyright, so if someone leaked it you could legally copy and distribute any parts which are AI generated but it wouldn’t invalidate copyright on the official binaries.
If all the code were AI generated (or enough of it to be able to fill in the blanks), you might be able to make a case that it’s legal to build and distribute binaries, but why would you bother distributing that slop?
Even if it were leaked, it would still likely be very difficult to prove that any one component was machine generated from a system trained on publicly accessible code.
Stallman: “Oh man, not like this.”
Counterpoint: how do you even prove that any part of the code was AI generated.
Also, i made a script years ago that algorithmically generates python code from user input. Is it now considered AI-generated too?
i made a script years ago that algorithmically generates python code from user input. Is it now considered AI-generated too?
No, because you created the generation algorithm. Any code it generates is yours.
While nobody created neural nets and back propagation
I believe you’re claiming that compiler authors own the software their compiler compiles which is clearly not true.
No way does op own the algorithmically created program based on user input
Computer output cannot be copyrighted, don’t focus on it being “AI”. It’s not quite so simple, there’s some nuance about how much human input is required. We’ll likely see something about that at some point in court. The frustrating thing is that a lot of this boils down to just speculation until it goes to court.
OP is obviously ignorant of how much tooling has already helped write boiler plate code.
Besides AI code is actually one of the things that’s harder to detect, compared to prose.
And all that said, AI is doing an amazing job writing a lot of the boilerplate TDD tests etc. To pretend otherwise is to ignore facts.
AI can actually write great code, but it needs an incredibly amount of tests wrapped around and a strict architecture that it’s forced to stick to. Yes, it’s far too happy sprinkling magic constants and repeat code, so it needs a considerable amount of support to clean that up … but it’s still vastly faster to write good code with an AI held on a short leash than it is to write good code by hand.
deleted by creator
deleted by creator
There is absolutely no way you wrote terabytes of code lmao.
deleted by creator
Uh, yes, that’s what they call a generative ai
By that same logic LLMs themselves (by now some AI bro had to vibe code something there) & their trained datapoints (which were on stolen data anyway) should be public domain.
What revolutionary force can legislate and enforce this?? Pls!?
By that same logic LLMs themselves (by now some AI bro had to vibe code something there)
I’m guessing LLMs are still really really bad at that kind of programming. The packaging of the LLM, sure.
& their trained datapoints
For legal purposes, it seems like the weights would be generated by the human-made training algorithm. I have no idea if that’s copyrightable under US law. The standard approach seems to be to keep them a trade secret and pretend there’s no espionage, though.
The packaging of the LLM, sure.
Yes, totally, but OP says a small bit affects “possibly the whole project” so I wanted to point out that includes prob AIs, Windows, etc too.
This reminds me of that scene in Breaking Bad where the two morons were talking about how if you ask an undercover cop if they’re cop they legally have to tell you the truth
So by that reasoning all Microsoft software is open source
Not that we’d want it, it’s horrendously bad, but still
The the greatest intellectual property theft and laundering scheme in human history
Windows is not even source-available. Windows XP is source-unintentionally-available thanks to a leak but there’s no AI loophole in that.
Does anyone know of any place to keep up with what people are doing with the XP leak? I vaguely remember some 4chan threads where people worked to get it compiling properly, and I think someone ported USB 3.0 support, but I lost track after that.
Not how copyright works. Adding something with creative height together with something without leaves the combined work with ownership only of the part with creative height with the rest unprotected.
(bots can not achieve creative height by definition in law)
How the hell did he arrive at the conclusion there was some sort of one-drop rule for non-protected works.
Just because the registration is blocked if you don’t specify which part is the result of human creativity, doesn’t mean the copyright on the part that is the result of human creativity is forfeit.
Copyright exists even before registration, registration just makes it easier to enforce. And nobody says you can’t just properly refile for registration of the part that is the result of human creativity.
Yeah, a lot of copyright law in the US is extremely forgiving towards creators making mistakes. For example, you can only file for damages after you register the copyright, but you can register after the damages. So like if I made a book, someone stole it and starting selling copies, I could register for a copyright afterwards. Which honestly is for the best. Everything you make inherently has copyright. This comment, once I click send, will be copyrighted. It would just senselessly create extra work for the government and small creators if everything needed to be registered to get the protections.
Edit: As an example of this, this is why many websites in their terms of use have something like “you give us the right to display your work” because, in some sense, they don’t have the right to do that unless you give them the right. Because you have a copyright on it. Displaying work over the web is a form of distribution.
That edit had confused so many users over the years. They think they are signing away rights to their copyrighted work by agreeing to the platform’s EULA, but the terms granting them license to freely store and distribute your work? That’s literally what you want their service to do because you’re posting it with the intention of the platform showing it to others!
Granted, companies are using user data for other purposes too, so that’s a problem, but I’ve seen so so many posts over the last couple decades of people complaining about EULAs that describe core site functions…
Is Windows FOSS now?
Ew, no, thank you, I don’t want it.
Didn’t sources leak multiple times
Yeah, and it was ghastly.
And they weren’t even infested with slop yet, just spaghetti…
That’s terrible news. There’s no way I want my code to be open source. Then other people would see just how much spaghetti you can have in a codebase and still have it run.
I think anyone forced to use Windows 11 in 2026 already knows that. Although there the term “run” is stretched to the limit.
