Mastodawn

chrismckee Mar 5

This case shows how Open Source will die. With anyone just being able to pipe existing code and tests through an LLM and claiming that to be "clean room" (which is hogwash) no licensing can protect your work from being accumulated and monetized by anyone. The commons are actively being shredded in front of our eyes.

https://github.com/chardet/chardet/pull/322

chardet 7.0: ground-up MIT-licensed rewrite by dan-blanchard · Pull Request #322 · chardet/chardet

Summary This PR is for a ground-up, MIT-licensed rewrite of chardet. It maintains API compatibility with chardet 5.x and 6.x, but with 27x improvements to detection speed, and highly accurate suppo...

GitHub

Cassandrich Mar 5

@tante And useful fools who claim to be stewards of FOSS are laying the groundwork and arguments for the industry to do this. 🖕 this guy.

Andrew Wigglesworth Mar 5

@dalias @tante Non-copyleft licenses like the "MIT licence" are now being used as an active threat to free software.

Aslak Raanes Mar 5

@ecadre @dalias @tante Even if the original license was MIT, you can't remove the name of the contributors. The license isn't actually the issue here, but that these people don't care about copyright when it suits them.

Andrew Wigglesworth Mar 5

@aslakr @dalias @tante OK, so when people "rewrite" GNU and other copy-left software and use permissive licences they are breaking MIT etc licences too.

could have just as well said "public domain." the instrument of the idiocy is not the root cause here so much as the willingness to pretend this process is legitimate

CC: @[email protected] @[email protected]

Cassandrich Mar 6

@khm @ecadre @tante Hot take: the "pretending this process is legitimate" of "AI-laundering FOSS to seize and enclose it" and "watching the dissolution of US democracy" are the same thing and the same awful people.

@tante Ugh. It's taken me awhile to identify my feelings about AI and I think it's essentially the 5 facets of grief. Often anger and denial, always depression, and even occasionally bargaining (with my manager) or acceptance (resignation, really).

AI feels like such a loss and the nonstop barrage of breathless enthusiasm from the media and folks at work feels truly bewildering and *isolating*. It's heartbreaking to see my industry (software dev) blithely ignoring the (environmental, human, social, economic) cost of their new toy.

@leathekd
I thought devs would be the first to reject this (as they too should take pride in their skill as humans.) I couldn't have been more wrong.
@tante

@ozzelot @leathekd @tante I think a disconnect for those of us who enjoy tinkering with computers at whatever level is that a lot of people are interested mostly in the money that historically computer jobs have brought in. Now, are they also systematically undermining their own future wages? Yes. Do they realize or think about this? Doubtful.

Now, I don't have an answer for why the fuck they're doing it in the context of open source codebases. Résumé padding?

@spacelogic @ozzelot @tante That may be right. I think there are those, like me, who enjoy the act of solving issues with code and there are those who are motivated differently. Some by money, for sure. Some might see solving the problem with AI to be more rewarding than solving it "by hand" ("look how productive I am, I wrote 10,000 lines of code today!"). I just happen to be someone who will program and tinker all through life because I find it enjoyable. Even if I wasn't, I would like to think that I would look at the rest of the consequences of AI and still turn away.

I feel lucky and grateful that my one talent in life happened to be in a field that was in demand and paid well. I hope my career doesn't go away (I have a family to feed) but I worry that it has been forever changed for the worse. Maybe when the bubble pops we'll find less AI stuffed into every aspect of life, but I would bet that software companies will still pay to keep AI around to "enhance productivity" or "democratize coding" or reduce headcount.

gloriouscow Mar 6

@spacelogic @ozzelot @leathekd @tante

We risk overanalyzing what might be something a simple as a dopamine hit.

Creators know it feels good to create things. Every programmer has experienced a little emotional high, however small, seeing their code run successfully, or smashing a difficult bug, or writing a complex function and having it work right off the bat.

And for many of us that is a feeling that has never ceased to be pleasurable even after decades of coding.

The danger I suppose is thinking any of us are fully immune to replacing the process, the effort and the art of coding and proceeding directly to the reward.

Insert prompt, see result, receive dopamine.

Does it matter, philosophically, if the result is now disconnected from your personal pride, effort, skills and knowledge, if it tricks your dumb meat brain into releasing those sweet sweet endorphins?

So much of modern civilization involves convincing ourselves that we are far more sophisticated animals than we really are, perhaps because our need for base physical desires are camouflaged by the indirect ways by which modern life requires us to go about satisfying them.

gloriouscow Mar 6

@spacelogic @ozzelot @leathekd @tante

There's a million ways to justify it - let's take the example of chardet - if you look at the number of commits involved, Dan Blanchard did indeed spend a lot of time instructing and directing Claude to complete the task.

To that end, maybe people discover that they have a knack for management. They can take pleasure in the successful planning they credit the result to. He can tell himself that not just anyone could have prompted Claude this way - his unique skills and experience were the key to his success, and so his satisfaction is justified, as his ownership of the creative process.

And that's even not entirely wrong, perhaps. People can insert a joke about middle-management here, but it's a skill that people apparently do enjoy and find fulfilling enough to make a career out of it.

Dan FitzGerald Mar 6

@ozzelot @leathekd @tante in a lot of cases we dont have the ability to. "Use AI or we fucking fire you" is pretty popular right now.

Jérôme Petazzoni Mar 6

@ozzelot @leathekd @tante the vast majority of us reject it en masse. If you look at the GitHub link posted in top, there is an overwhelming ratio of people being unhappy about that behavior. But there is still a noisy minority embracing it, for sure.

@ozzelot @leathekd @tante Every single organization’s IT that I witnessed was pathologically dysfunctional. Vast majority of devs I worked with scoffed at formal verification and just went with the feel (vibes). I sat speechless as a CTO laid claim that the API client was broken because they were getting a „no route to host” error.

AI is a logical continuation of all that. It fits like a washed condom.

Mike Anderson Mar 5

@tante although it's deeply offensive and morally wrong, I don't think that we need panic about open source for this reason. It was always difficult to detect if a company took a copyleft product and parasitised it, and plenty of healthy open source projects have been under MIT-style licenses for decades.

@mspcommentary @tante However one reason I put things under MIT licenses was I believed the MIT license would be obeyed. If my only option were to make everything effectively public domain, I'm not sure I would have open sourced so many things.

Mike Anderson Mar 5

@mcc @tante totally, and I put things under the GPL to try to keep them open. I am horrified, frankly, that the LLMs are openly stealing intellectual property and governments seem to be totally OK with it. I just think that the real strength of open source comes from combining our efforts to make things that we all benefit from.

Mason Loring Bliss Mar 5

@tante Good discussion of the topic here:

https://softwareengineering.stackexchange.com/questions/260347/ship-of-theseus-applied-to-gpl-can-i-relicense-my-program-if-i-replace-all-of#260357

Technically, the chardet code is now illegal to copyright.

Also, because there haven't been any attempts at determining the source of every part of the code, it is unknown if there was any additional copyright violation.

Bogdan Buduroiu Mar 5

@tante oh can you cut it with the arm flailing outrage baiting, these statements are intellectually lazy and imo dishonest.

Yes, this is a violation of GPL, no, it's not the death of open source / shredding of the commons, because open source has been dead for decades:

Onyx has for years refused to release the kernel source code for their Linux-based e-readers despite GPLv2 requiring it, with users and communities begging for compliance across multiple forum threads spanning years, with no enforcement action taken. Their answer: "we won't open source because of anti-China sentiment, sorry".

Your "open" phone has more proprietary blobs than open ones.

People, despite their moralisation on the topic, don't care for open source, and the Jia Tan/xz vuln shows it.

degenerating degenerate Mar 5

@budududuroiu @tante I agree the chicken little thing is wrong (and getting old).

But why is it a "violation of the GPL"? When Toybox (0 clause BSD-licensed) was written from scratch to compete with Busybox (GPL2) years ago, before AI, the sky did not fall either. Doing it first and sticking a GPL License on it did not put an electric fence around the whole idea.

If the AI also followed eg the APIs and filled in its own code, what's even the problem?

Bogdan Buduroiu Mar 5

@hopeless @tante also Busybox didn't die as a result, contrary to what tante wants us to believe about the entirety of open source lmao

@tante On the other hand, anyone with "shared source" versions of windows can now easily make a clean room GPL windows version happen. If this result stands, that is.

Minecraft now ships with unobfuscated Java code, thinking about that...

Ray McCarthy Mar 5

@maswan @tante
Well, it's not clean room and it's nonsense.

@raymaccarthy @tante That's up to the courts to decide, I guess.

Will be interesting also for, say, Disney to do "clean room" adaptations of books still in copyright, so they don't have to pay anything to authors.

Ray McCarthy Mar 5

@maswan @tante
Disney have a history of doing that anyway:
See Bambi and later royalties due when they bought Lucas.

Also leaves them open to having no copyright on the "new" work.

@tante This code actually is in the public domain now, if it was ever legal to begin with.

The MIT license is just some stupid decoration the user of the AI (or the AI itself) decided to stick on the uncopywritable output the AI produced.

In the USA anyway. Until an act of congress. I actually expect it to be pretty fast.

Ray McCarthy Mar 5

@crazyeddie @tante
Anything produced by AI is PD: Can't be copyrighted. Seems to be the recent USA supreme court decision?

@raymaccarthy @tante Yep.

I do expect it to get changed really fast by congress. Congress only appears to be in deadlock. They'll have no problem uniting to fix this. For now though the result seems to be that the output of AI can't be copyrighted and it'll take an act of congress to change that. I don't think even an EO can change this, though won't stop him from trying nor from SCROTUS to just turn around and say, yep...here's the spaghetti wordsmithing that says he can.

Tiota Sram Mar 5

@crazyeddie @raymaccarthy @tante the decision isn't that simple. Works produced by a human with the "help" of AI are still copyrightable, which the author here claims is the case. What exact degree of human vs AI involvement is the line between copyrightable and not is an interesting and unsettled question, but it's not at all clear that the human involvement here doesn't meet that bar.

Ray McCarthy Mar 5

@tiotasram @crazyeddie @tante
Maybe, and it depends on country and courts.

It's stupid.

@tante surprise surprise he works in fintech.

@tante I have heard people talking about this but what I don't understand is what license the code was under before the "rewrite". The project's own self-description says "chardet 7.0 is a ground-up, MIT-licensed rewrite of chardet" which sounds like something a LLM would write but doesn't tell me much (And the PR is big enough to break Github's PR display feature so it's a little hard to figure out what the project looked like before the +14526 -546715 patch.)

@mcc Before the Claude "clean room" reimplementation chardet was licensed under LGPL.
So the dude used claude (which was probably trained on chardet/LGPL) to generate a new version of chardet with the same API etc but put it under MIT license.

@tante Thank you. And do I understand correctly that administratively the MIT rewrite* is "the same library", e.g. the maintainer flattened their own repository and hosted the new* thing at the same github address, same pypi address, same readthedocs address? I hadn't used Chardet previously and search engines point only to this same project.

* "so to speak"

@mcc yes. They jumped from 6.0.0 (LGPL) with a huge merge request that changed everything to the Claude generated version and relicensed it. So it is in the same tree. Same name/canonical URL and everything. So the "clean room" argument is at least softened by putting it in direct succession

@tante Okay. It's clear now, Tthanks.

(Presumably of course the single maintainer did not write all of that code and anyone who contributed under LGPL would have expected consent before relicensing.)

@mcc yes, there are contributors who contributed under LGPL especially Mark Pilgrim who started the whole project under that license.

SomeVeganCheeseIsOk Mar 5

@tante @mcc does the fact that AI content can't be copyrighted impact this at all?

@SomeVeganCheeseIsOk @mcc "AI content can't be copyrighted" is a bit of an oversimplification. TBH for the matter at hand it is relevant only in the sense that Claude surely was trained on chardet so it supports the "derivative" argument a bit.

@tante Time to dust off https://gpl-violations.org?

The gpl-violations.org project

The gpl-violations.org project tries to raise public awareness about past and present infringing use(r)s of GPL licensed software. The ultimate goal is to make companies engaging in the distribution o

gpl-violations.org

@tante
just some quick checks to see what it is...

"chardet 7.0 is a ground-up, MIT-licensed rewrite of chardet" [links to itself]

documentation: first thing shows a test of an English sentence and it thinks it's Spanish.

Original author complains about the re-licensing:
https://github.com/chardet/chardet/issues/327

I might need more popcorn

No right to relicense this project · Issue #327 · chardet/chardet

Hi, I'm Mark Pilgrim. You may remember me from such classics as "Dive Into Python" and "Universal Character Encoding Detector." I am the original author of chardet. First off, I would like to thank...

GitHub

Chris Real Mar 5

Lenovo just rejected years of Microsoft's incompetent monopoly for open source systems, that are more usable from a blunt capitalist perspective.

Your concerns are trivial and easily dealt with.

A monopoly who wants to rent an invasive and bloated system to people who have been FED UP with it for DECADES?

This case shows the death of incompetent monopolists, who could never prevail in a world without government-assisted monopolies.

Their anger is just the hissy fit of privileged nepo's* who can't cope with competition in the real world.

*(Gates' parents were millionaires at the time he started claiming ownership of 'DOS'.

Which he BOUGHT from another programmer.)

https://youtube.com/shorts/S8zIEqsXwq4

Bill Gates Did Not Write the Original MS- DOS Himself #debunkingmyths

YouTube

Pete / Syllopsium Mar 5

@tante Whilst this is an interesting and popcorn worthy thread, and has many possible implications I would point out you mean GPL, you do not mean Open Source. Plenty of BSD code out there that anyone can do whatever they want with.

It is also nothing to do about monetization - it's entirely possible to monetize GPL code, particularly LGPL.

I'd point out it's nothing new. Take Unihertz, for instance, who have repeatedly released smart phones based on GPL code and not released their modifications. Not many people shouting about it, and what has happened, exactly?

Waitman Gobble Mar 6

They can do as long as they follow the terms of the license. People have been improperly using licensed code way before AI. Just adding a BSD license is not techically legit nor is saying it's BSD licensed.

Router mfgs are horrid about breaking the license. I was looking at admin of one today that says it contains GPL code with a link to GPL license on their site. Where is the code?

craignicol Mar 5

@tante @PeteKirkham given how they were trained, it's impossible for any of the current generation of LLMs to produce a clean room implementation of anything.

Grow Fediverse Mar 6

@tante No human, no copyright/patent ... and probably as a consequence, no licensing either.

FLOSSbOxIN Mar 6

@tante
https://github.com/chardet/chardet/issues/327

No right to relicense this project · Issue #327 · chardet/chardet

Hi, I'm Mark Pilgrim. You may remember me from such classics as "Dive Into Python" and "Universal Character Encoding Detector." I am the original author of chardet. First off, I would like to thank...

GitHub

Robert Thau Mar 6

@tante But this is how open source itself got started -- RMS implementing GNU Emacs incorporating large chunks of Gosling Emacs's code (against terms of its license), later rewriting but preserving the interfaces, and writing gcc having access to the source of lcc, and very likely pcc. (Yes, Gosling rewrote RMS's earlier work.)

"Clean room" has never been a requirement of copyright law. It's a defensive tactic going above and beyond legal requirements to fend off particularly vicious litigants.

nojhan à la double douleur vaudoue Mar 6

@tante They violate copyright so that they can violate copyright even more…

Thomas Depierre Mar 6

@tante i don't like it but... I think it is pretty separated from open source dying.

Open source exist because i want to put my stuff up available without consequences. That crap doesn't change it.

@Di4na @tante I'd argue open source is about more than just the right/ability to *host* my code/software.

To me, it necessarily must also mean that there's a community out there that can maintain good projects (e.g. the Linux Kernel) and keep them from being swallowed by proprietary aquisitions¹.

Open source requires a legal basis that can help a project maintain its status and license. Contributors shouldn't have to worry about possible re-licensing attempts 10 years down the line.

A contributor should have reasonable assurances that the project they contribute to isn't going to somehow utilize those contributions towards a differently licensed², AI-sloppified version of what you wrote, with your copyright ownership conveniently stripped of you.

---

¹ Wouldn't Android just love to move all development of Linux over to a closed-source repo? Or to have their own specially modified version of it by forking an MIT-licensed version, instead of needing to comply with the GPL license?

² differently licensed using a new license that is incompatible with the original. Especially damning with LGPL -> MIT, where the terms are massively weakened in the re-license.

Thomas Depierre Mar 6

@riverpunk @tante I understand you.

I also... How to tell you. That community part? It is incredibly rare. On average and even on p95, OpenSource is just one person.

I get that it is not the ethos and ideology you may have. That is fair and good. Keep up the fight.

But it is also not out current reality

Stormbringer Mar 6

AI code can't have copyright.

https://pluralistic.net/2026/03/03/its-a-trap-2/

Pluralistic: Supreme Court saves artists from AI (03 Mar 2026) – Pluralistic: Daily links from Cory Doctorow

@laryllan @tante important to note here:

the AI outputted code need not be copyrightable in order to *remove* the old license. The relevant fact for that would be to see if the rewrite was "clean" or not (which as stated above, it's of course not, so this is considered a derivative work, thus via LGPL still owned by original authors).

As for applying a new MIT license, yeah this is hairy. The AI prompters might be doing some original coding of their own beyond just copy-pasting, in which case they retain a minimal amount of copyright overall (but most of it is public domain). So like, yeah they can't really be putting MIT unless they're doing tangible work of their own on top of AI boilerplate.

Code licenses tend to be designed to recognize all contributions, even small ones, so there's probably some room for debate, depending on their workflow.

But again, not that that matters, cuz removing the LGPL wasn't legal in the first place, since this wasn't a clean, ground up implementation.

(FYI I haven't researched this case too much, I've just read some of the comments in the thread lol)

@tante
This is why I am going back to retro. Old ways, before git repos, and all that. Sharing code the old ways

Rob Landley Mar 6

@tante Source code wasn't copyrightable before 1976 and binaries weren't until 1983. The complete elimination of copyright in software would make open source the default again.

LLMs suck but "copyleft" was always a kludge. Might need to push the other way instead. (Stallman celebrating how the DMCA gave GPLv2 more teeth squicked me out at the time...)