[COPYRIGHT] Take down this stupid repo: https://github.com/ultraworkers/claw-code

Is this supposed to read like satire? Many of these complaints and arguments are exactly the same arguments levied at LLM programming assistants in the first place.

What they’re doing is taking other people’s hard-earned work, repackaging it, and pretending it’s their own so they can profit from it. How is that even remotely ethical?

This is more or less exactly what it's commonly claimed that AI agents do to FOSS code in the first place.

Anyway, it's not clear that that repository is copyright infringement. I don't think there's any legal precedent as to whether how that was developed counts as a derivative product or not, given that it's a "clean room" implementation on code that's probably not in the agents' training sets. I think it's an open question.

This is exactly what people mean when they say it’s dangerous when someone with shallow, half-baked knowledge starts forming strong opinions.

Comparing open source code that was intentionally made public to corporate intellectual property that is clearly owned is not something you can just joke about and move on from. That’s not a harmless take. That’s a complete lack of understanding of how this actually works. Even before AI, open source license violations were already happening everywhere. Companies didn’t chase every single case, and they didn’t need to.

Let me put it simply. If you spent five years building software and it got leaked, and someone grabbed it, ran it through AI, rewrote it in another language, and then claimed it as their own, who owns it? If you seriously believe that running something through AI magically turns it into a brand new copyrighted work, then you are basically arguing that copyright should not exist at all. That kind of thinking is insane. If that is your logic, then just delete copyright entirely and be done with it.

What’s even worse is watching people encourage this behavior. This is the level of someone with no real industry experience, like a clueless new grad. Instead of experienced engineers stepping in, calling it out, and guiding people in the right direction, everyone is cheering it on because they are blinded by the idea of getting something for free. It’s honestly embarrassing.

Yes, there is value in learning from Claude Code. That’s obvious. But if you have any common sense, you should be asking whether this is actually healthy for the ecosystem. What is going to happen is tools like Gemini or Codex, which do not have the same level of capability, will start piggybacking on Claude Code’s agent logic. That leads to endless knockoffs, and it will directly impact Anthropic’s revenue. On top of that, the marketing damage from publicly exposing this is going to be massive.

Is this really the “free” outcome you wanted? Because in the end, users are the ones who will pay for it.

So, yea. This isn’t satire. It’s just a basic distinction that you’re completely missing.

The argument you’re making about LLMs and FOSS is not equivalent here, and pretending it is just shows a lack of understanding of both copyright and how these systems are actually used.

Open source is explicitly licensed for reuse under defined terms. Corporate proprietary code is not. That’s the entire point. Collapsing those two into the same bucket is not a clever argument, it’s just very very stupidly wrong.

And no, this is not some abstract philosophical debate about AI. This is about taking non-public, proprietary work, reconstructing it, and distributing it in a way that clearly mirrors the original system. Calling that “clean room” without any real scrutiny doesn’t magically make it legitimate.

Also, saying “there’s no legal precedent” is not the defense you think it is. It just means the risk is even higher, not lower. Acting like that somehow makes it safe is naive at best.

What’s actually happening here is people trying to rationalize behavior that would be obviously unacceptable in any other context. The only difference is that now it’s dressed up with AI, so suddenly everyone pretends the rules don’t apply.

Copyright doesn’t require exact copying to apply. Structural similarity alone can be enough, and if you’re really in the U.S., you should already know that. Or maybe you just don’t.

What you’re arguing is equivalent to copying someone else’s work, rewriting it in your own handwriting style, and pretending it’s yours. That’s not clever, it’s just stupidly WRONG.

Anthropic is already issuing takedowns across ten thousands of repos, and the people who think everything is “free” are about to learn otherwise.

If a colleague leaked confidential information, they’d be fired immediately and held accountable. But somehow people are out here defending clueless, immature students with no real involvement. It’s ridiculous how backwards this has all become.

I'm not going to get way into the weeds with this one, and I don't believe this is really the place for it, but I'll address a few points, and I'll also note that I'm not some "clueless new grad", but have been writing software professionally and for open source for 15 years, and paying very close attention to licenses, copyright, and open source. I'm not a lawyer, but I have spent at least a decade reading licenses and paying attention to licensing lawsuits, and have a pretty good understanding of their functioning. I'll also note that I am a paying customer on a Claude Max subscription, so I'm not some anti-AI activist.

Comparing open source code that was intentionally made public to corporate intellectual property that is clearly owned...

...is sane and normal to do, given that both have equal protection by copyright. It's largely tangential to the point, but training a LLM on FOSS and generating derivative code without having to respect the origin's copyright should have no less protection than training on a proprietary codebase and doing the same. Corporate code deserves no more or less protection under copyright than FOSS. You speak as if it is somehow protected by stricter laws. It is not. It is protected by better-paid lawyers.

If I spent decades building software and it got leaked, one person read the sources and wrote specifications, and another person read the specifications and implemented it, that would be a clean room reverse engineering process, which ReactOS and Wine follow when developing their software. That repository used the same process. This was not laundered through an LLM, but one LLM read it and wrote a specification, and another implemented the specification. Also an interesting note: Claude Code is largely written by Claude itself. Who owns that copyright?

If this is suspect, legally, then so is all the output that Claude Code generates in the first place. I never argued that this is in the clear, or that it is not. I argued that the implications of the legality of that Claw Code project and all of the output of Claude are pretty directly linked.

And again:

If you seriously believe that running something through AI magically turns it into a brand new copyrighted work, then you are basically arguing that copyright should not exist at all. That kind of thinking is insane. If that is your logic, then just delete copyright entirely and be done with it.

This reads exactly like an anti-LLM argument. That's exactly what LLMs are doing already to FOSS code. I use LLMs at work, but I'm not going to delude myself about the shaky legal and ethical grounds they already stand on.

So who exactly is executing Claude Code?

You’re trying very hard to sound nuanced here, but you’re collapsing multiple completely different legal and technical concepts into one argument and pretending they’re equivalent. They’re not.

Saying “FOSS and proprietary code have equal copyright protection” is technically true in the most superficial sense, but completely irrelevant to the situation. The entire point is not whether they are protected, but under what terms they are allowed to be used, reproduced, and redistributed. Open source is explicitly licensed for reuse. Proprietary code is not. That distinction is not subtle.

Your “clean room” comparison is also doing a lot of heavy lifting without actually holding up. ReactOS and Wine operate under extremely strict processes with legal oversight, separation of teams, and documented compliance. What you’re describing here is nowhere near that standard. Calling it “clean room” does not make it one. It just shows you’re borrowing a term without respecting the rigor behind it.

And the LLM comparison is a deflection, not an argument. Yes, there are ongoing debates about training on FOSS and generated outputs. That does not magically legitimize reconstructing proprietary systems and redistributing them. Two legally ambiguous areas do not cancel each other out.

“Claude writes Claude Code, so who owns it?” is also not the clever point you think it is. That is just STUPID. Ownership is determined by the entity controlling the system and the terms under which it operates, not by anthropomorphizing the model. This has already been addressed in multiple legal frameworks.

What you’re really doing is taking unresolved questions around LLMs and using them as a blanket justification for behavior that would be obviously unacceptable in any non-AI context. That’s not a strong argument. It’s rationalization.

And leaning on “there’s no clear legal precedent” as a defense just makes this worse, not better. It means you’re operating in a high-risk area and choosing to ignore it. That’s not informed, it’s reckless.

You clearly have experience, but right now you’re using it to blur boundaries that actually matter. That’s the problem.

You idiots who haven't even gotten a job yet, let's hear your story. See if you are acting so arrogantly that it is lowering the status of your country (probably North Korea, right?)

@Yeachan-Heo @sigridjineth @code-yeongyu

Oh yea, S. Korea 😂

Open source is explicitly licensed for reuse. Proprietary code is not. That distinction is not subtle.

Open source is explicitly licensed for reuse under the explicit terms of the license. If you don't follow those terms explicitly, including attribution and share-alike terms, you are violating the license and infringing on copyright, just as any infringement on proprietary software. The situations are not as different as you seem to think. There's no real legal distinction between FOSS and proprietary software as categories, just copyright and license terms.

I don't have anything to do with the Claw Code repository and have no interest in "defense" of it at all. I'm interested in the legal and copyright questions, not any particular repository. I'm not "operating in a high-risk area and choosing to ignore it", I'm pointing out that it's not as clear a situation as you pretend. You're getting weirdly defensive and insulting. Why are you taking this so personally? Sock puppet account?

Claw Code was created from the ground up, not directly stolen from source. So really, you have nothing to stand on when it comes to copyright claims. And you know copyright only protects a specific expression, not ideas or functionality. The way you express your rant is honestly quite comical and rather dramatic; you act as if it is your very own source code. The only impact would be on Anthropic, which I'm sure they can recover from. They are a multi-billion-dollar company, after all. It's only an SDK that got leaked. And it's basically well-written code with some clever prompting. I could think of one or two more cases that are far more impacting when it comes to reverse engineering, and this isn't even that big of a scandal. For someone who has supposedly been in the industry this long, you sure think you know a whole lot.
Taking it down isn't going to serve much of anything, not when the source is out, and repos will keep popping up here and there.
And finally, I can stop paying for the Claude Code plan and instead make use of Claude through the API when I need it for complex tasks, and use something like GPT-4 or something cheaper instead. Now, that is value, and a true benefit. I'm sure there are plenty of mishaps people have made that others would've happily taken advantage of, and this is no different.

Closing for now — inactive for too long. Please open a new issue if this is still relevant.

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.