TL;DR: The advent of AI based, LLM coding applications like Anthropic’s Claude and ChatGPT have prompted maintainers to experiment with integrating LLM contributions into open source codebases.
This is a fast path to open source irrelevancy, since the US copyright office has deemed LLM outputs to be uncopyrightable. This means that as more uncopyrightable LLM outputs are integrated into nominally open source codebases, value leaks out of the project, since the open source licences are not operative on public domain code.
That means that the public domain, AI generated code can be reused without attribution, and in the case of copyleft licences - can even be used in closed source projects.
Counterpoint: find all corporate vibe coded IP and make everything public domain
deleted by creator
That seems pretty good to me? I hate LLMs, but this policy is basically “if it’s obviously LLM garbage or you don’t understand it, it will be rejected” and I’m not sure it’s practical to do better.
People will use LLMs behind the scenes, but if they are able to write a coherent justification with clear understanding of the code, receive feedback from devs and rework it, as well as submitting code that is well structured etc. it’s not really any different than any other PR.
Except for the fact that it is public domain and not protected by the open source license that the code is ostensibly submitted under.
How are the devs or anyone else supposed to tell that though, if all the LLM trappings are absent?
That seems pretty good to me? I hate LLMs, but this policy is basically “if it’s obviously LLM garbage or you don’t understand it, it will be rejected” and I’m not sure it’s practical to do better.
You did not read/watch the content of this post.
I read the post, but as I mentioned elsewhere, how are devs (or malicious commercial thieves looking for public domain code) supposed to detect this code is an LLM creation when all of the obvious signs they mention are stripped?
A ban on people using an LLM in secret is unenforceable and the code output can be indistinguishable from a human’s, especially when a real human that understands the change is there to baby it and write commit messages etc.
So copyright underpins every license type? I didn’t realize that. I assumed they were separate and focused on different contexts.
All works created by a person are copyright by default, so people need to release their works to allow others to build on it or use it (except for the limited uses allowed by fair use). Like-minded people have come up with various licenses that allow people to release their works in ways that people prefer.
And AI code is unable to be licensed? Because there is no copyright? I can’t just put a restrictive license on it that says only people who meet an impossible criteria can use it? Thus blocking use?
How does this apply to software made by, say, Anthropic? They proudly say Claude Code is written by AI. If it can’t be copywritten, or licensed, then it’s just a matter of figuring out how to acquire a copy of the source code, and you could do whatever with it. Right?
How does this apply to software made by, say, Anthropic? They proudly say Claude Code is written by AI. If it can’t be copywritten, or licensed, then it’s just a matter of figuring out how to acquire a copy of the source code, and you could do whatever with it. Right?
If you were on Mastodon last week when the Claude source code was released (by Claude, accidentally), people were joking about how Anthropic was trying to use the DMCA to get the source removed from websites – even though clearly, copyrights don’t apply, since the code is clearly in the public domain.
If the LLM wrote the code, it is uncopyrightable.
Yes, but can it be licensed. Why is my question so confusing? Is copyright a prerequisite to a license?
What do you mean? What do you think the license is for? If I own the copyright to something, only I can use it but I can allow others to use it by giving them a license. If something is not copyrighted, anyone can use it because no one controls the rights.
I guess it was just as simple as you say, I was just looking for a “yeah, that’s how it is.” I guess. I was having a conversation, looking for a clear reflection that I was correct or not.
But also, I feel like there must be room for exploration here, since mixing AI generated with hand written is a thing. And there must be something around the facilitation/collaboration part. It feels like it’s not just as simple as free code that is controlled via access to the source.
I totally understand that purely generated output is not copyrightable. That’s clear. But I feel like there must be grey area yet to be litigated.
Can I legally reverse engineer AI generated software? Does it matter what modules are hand written and what ones are not when I reverse engineer it? How is this accounted for in licenses? Terms of Service?
Can you even put terms and conditions on this supposed public domain copyright free compiled software product? Etc. etc.
Is the compiled version even different than the raw AI generated source code in its ability to be licensed?
How do you prove code is AI generated? How do you pick it apart when it’s only ai augmented?
This is why I’m asking the questions I am. It just doesn’t feel like it all hangs together clearly.
What rights does one have to AI generated code? Be it compiled or source. It’s surely not just communal.
Can I legally reverse engineer AI generated software?
If you have the source, why would you need to?
Can you even put terms and conditions on this supposed public domain copyright free compiled software product?
You can put terms on anything, but you can’t protect the underlying asset if someone breaks your terms. Think of the code produced by Grsecruity that they put behind a paywall – people were free to release the code (since it was licensed as open source as a derivative work), but obviously Grsecruity was able to discontinue their agreement with their clients who would do so.
Is the compiled version even different than the raw AI generated source code in its ability to be licensed?
People aren’t generally licensing compiled binaries as open source, since you can’t produce derivative works from them. But I think that if there is no copyright protection for the work, compiling it doesn’t change the copyrightability. Curious what you think.
What rights does one have to AI generated code? Be it compiled or source. It’s surely not just communal.
Why is that surely the case? It is public domain - that is the most “communal” you can get for copyright.
