Relicensing with AI-Assisted Rewrite • Hacker News Board

Re: Relicensing with AI-Assisted Rewrite

#231

Post by orthoxerox » Thu, Mar 05, 2026, 2:19 PM UTC

foltik wrote: post #179

shevy-java wrote: post #94

> In traditional software law, a “clean room” rewrite requires two teams So, I dislike AI and wish it would disappear, BUT! The argument is strange here, because ... how can a2mark ensure that AI did NOT do a clean-room conforming rewrite? Because I think in theory AI can do precisely this; you just need to make sure that the model used does that too. And this can be verified, in theory. So I don't fully understand a…

Turns out there’s no need to speculate. Someone pointed out on GH [0] that the AI was literally prompted to copy the existing code: > *Context:* The registry maps every supported encoding to its metadata. Era assignments MUST match chardet 6.0.0's `chardet/metadata/charsets.py` at https://raw.githubusercontent.com/chardet/chardet/f0676c0d6a... > Fetch that file and use it as the authoritative reference for which enco…

That's data, not code.

Re: Relicensing with AI-Assisted Rewrite

#232

Post by vbarrielle » Thu, Mar 05, 2026, 2:21 PM UTC

abrookewood wrote: post #71

_dwt wrote: post #40

Earlier quoted context omitted.

Yes, this is the reason I've completely stopped releasing any open-source projects. I'm discovering that newer models are somewhat capable of reverse-engineering even compiled WebAssembly, etc. too, so I can feel a sort of "dark forest theory" taking hold. Why publish anything - open or closed - to be ripped off at negligible marginal cost?

Why does it matter if it is 'ripped off' if you released it as open source anyway? I get that you might want to impose a particular licence, but is that the only reason?

Even the most permissive open source licenses such as MIT require attribution. Releasing as open source would therefore benefit the author through publicity. Bein able to say that you're the author of library X, used by megacorp Y with great success, is a good selling point in a job interview.

LLM ripping off open source code removes that.

Re: Relicensing with AI-Assisted Rewrite

#233

Post by bandrami » Thu, Mar 05, 2026, 2:22 PM UTC

thousand_nights wrote: post #205

danlitt wrote: post #134

I am pretty sure this article is predicated on a misunderstanding of what a "clean room" implementation means. It does not mean "as long as you never read the original code, whatever you write is yours". If you had a hermetically sealed code base that just happened to coincide line for line with the codebase for GCC, it would still be a copy. Traditionally, a human-driven clean room implementation would have a vanish…

the whole concept of a "clean room" implementation sounds completely absurd. a bunch of people get together, rewrite something while making a pinky promise not to look at the original source code guaranteeing the premise is basically impossible, it sounds like some legal jester dance done to entertain the already absurd existing copyright laws

In the archetypal example IBM (or whoever it was) had to make sure the two engineering teams were never in the cafeteria together at the same time

Re: Relicensing with AI-Assisted Rewrite

#234

Post by jacquesm » Thu, Mar 05, 2026, 2:22 PM UTC

SiempreViernes wrote: post #206

jacquesm wrote: post #173

Earlier quoted context omitted.

> it isn't clear how/if llm is different from the brain It's very clear: the one is a box full of electronics, the other is part of the central nervous system of a human being. > but we all have training by looking at copywrited source code at some time. That may be so, but not usually the copyrighted source code that we are trying to reproduce. And that's the bit that matters. You can attempt to whitewash it but at…

> but we all have training by looking at copywrited[sic] source code at some time. The single word "training" is here being used to describe two very different processes; what an LLM does with text during training is at basically every step fundamentally distinct from what a human does with text. Word embedding and gradient descent just aren't anything at all like reading text!

Indeed, but that's just a misdirection. We don't actually know how a human brain learns, so it is hard to base any kind of legal definition on that difference. Obviously there are massive differences but what those differences are is something you can debate just about forever.

I have a lot of music in my head that I've listened to for decades. I could probably replicate it note-for-note given the right gear and enough time. But that would not make any of my output copyrightable works. But if I doodle for three minutes on the piano, even if it is going to be terrible that is an original work.

Re: Relicensing with AI-Assisted Rewrite

#235

Post by ndriscoll » Thu, Mar 05, 2026, 2:23 PM UTC

jacquesm wrote: post #228

ndriscoll wrote: post #222

Earlier quoted context omitted.

How exactly is it different? All the model itself is is a probability distribution for next token given input, fitted to a giant corpus. i.e. a description of statistical properties. On its own it doesn't even "do" anything, but even if you wrap that in a text generator and feed it literal gcc source code fragments as input context, it will quickly diverge. Because it's not a copy of gcc. It doesn't contain a copy of…

There already have been multiple documented cases of LLMs spitting out fairly large chunks of the input corpus. There have been some experiments to get it to replicate the entirety of 'Moby Dick' with some success for one model but less success with others most likely due to output filtering to prevent the generation of such texts, but that doesn't mean they're not in there in some form. And how could they not be, it…

Are you referring to this?

https://osyuksel.github.io/blog/reconstructing-moby-dick-llm...

I see a test where one model managed to 85% reproduce a paragraph given 3 input paragraphs under 50% of the time.

So it can't even produce 1 paragraph given 3 as input, and it can't even get close half the time.

"Contains Moby Dick" would be something like you give it the first paragraph and it produces the rest of the book. What we have here instead is a statistical model that when given passages can do an okay job at predicting a sentence or two, but otherwise quickly diverges.

Re: Relicensing with AI-Assisted Rewrite

#236

Post by hrmtst93837 » Thu, Mar 05, 2026, 2:25 PM UTC

anilgulecha wrote: post #3

This is precedent setting. In this case the rewrite was in same language, but if there's a python GPL project, and it's tests (spec) were used to rewrite specs in rust, and then an implementation in rust, can the second project be legally MIT, or any other? If yes, this in a sense allows a path around GPL requirements. Linux's MIT version would be out in the next 1-2 years.

[flagged]

Re: Relicensing with AI-Assisted Rewrite

#237

Post by kevin_thibedeau » Thu, Mar 05, 2026, 2:27 PM UTC

brians wrote: post #186

danlitt wrote: post #134

I am pretty sure this article is predicated on a misunderstanding of what a "clean room" implementation means. It does not mean "as long as you never read the original code, whatever you write is yours". If you had a hermetically sealed code base that just happened to coincide line for line with the codebase for GCC, it would still be a copy. Traditionally, a human-driven clean room implementation would have a vanish…

I do not agree with your interpretation of copyright law. It does ban copies : there has to be information flow from the original to the copy for it to be a "copy." Spontaneous generation of the same content is often taken by the courts to be a sign that it's purely functional, derived from requirements by mathematical laws. Patent law is different and doesn't rely on information flow in the same way.

Derivative works can also run afoul of copyright. An LLM trained on a corpus of copyrighted code is creating derivative works no matter how obscure the process is.

Re: Relicensing with AI-Assisted Rewrite

#238

Post by satvikpendem » Thu, Mar 05, 2026, 2:28 PM UTC

pornel wrote: post #185

Generative AI changed the equation so much that our existing copyright laws are simply out of date. Even copyright laws with provisions for machine learning were written when that meant tangential things like ranking algorithms or training of task-specific models that couldn't directly compete with all of their source material. For code it also completely changes where the human-provided value is. Copyright protects…

Honestly, good. Copyright and IP law in general have been so twisted by corporations that only they benefit now, see Mickey Mouse laws by Disney for example, or patenting obvious things like Nintendo or even just patent trolling in general.

Re: Relicensing with AI-Assisted Rewrite

#239

Post by orthoxerox » Thu, Mar 05, 2026, 2:28 PM UTC

duskdozer wrote: post #230

orthoxerox wrote: post #223

Earlier quoted context omitted.

So, if these commits were private and squashed together before 7.0 was published there would be no violation?

They would be concealing the violation.

Consider TCC relicensing. They identified the files touched by contributors that wanted to keep the GPL license and reimplemented them. No team A/team B clean room approach used. The same happened here, but at a different scale. All files now have a new author and this author is free to change the license of his work.

Re: Relicensing with AI-Assisted Rewrite

#240

Post by bandrami » Thu, Mar 05, 2026, 2:30 PM UTC

pavlov wrote: post #99

nairboon wrote: post #9

That code is still LGPL, it doesn't matter what some release engineer writes in the release notes on Github. All original authors and copyright holders must have explicitly agreed to relicense under a different license, otherwise the code stays LGPL licensed. Also the mentioned SCOTUS decision is concerned with authorship of generative AI products. That's very different of this case. Here we're talking about a tool t…

If anything, the SCOTUS decision would seem to imply that generative AI transformations produce no additional creative contribution and therefore the original copyright holder has all rights to any derived AI works. (IANAL)

That's a reasonable theory though it's stuck with the problem that any model will by its training be derivative of codebases that have incompatible licenses, and that in fact every single use of an LLM is therefore illegal (or at least tortious).