Dan Blanchard, maintainer of Python's "chardet" library, used Claude to rewrite the entire project's codebase so that he can switch the license from LGPL to MIT.

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

I highly doubt that this is legal, but who the fuck cares these days anyway, right?

If I was a contributor to that project, I'd tell him in no vague words what I think about shit like that. "No bro, it's totally not relicensing your code bro, this is totally new code bro!"

via https://chaos.social/@Foxboron/116170859737134271

"The test data has also been moved to a separate repo to help prevent any licensing issues"

So you've used the LGPL-licensed tests provided by contributors as a spec for Claude?

And of course Claude wasn't trained on the existing chardet code, nu-uh.

Like, usually when trying to free code from copyright or NDAs, it's an elaborate process, with the authors having to guarantee to never even have looked at the original code.

Did you add "pretend you didn't see the original" to the prompt? lol.

Damn I hope we spend 2027 to 2037 suing the whole LLM industry into the fucking ground.

Every software project built using license-violating training data, every piece of music made by stealing from thousands of artists simultaneously, every picture created by ripping off people who barely have enough to make a living.

Sue them all for copyright violations.

I don't care if we need twice as many lawyers to do that.

Or, option B, abandon copyright altogether. Both works for me.

@scy wird das nicht für WINE total witzig wenn da jemand zeug mit Copilot bastelt das dann ggf. Auch mal die original Windows Repos gesehen hat?

Hm... Ob wohl jemand original Windows Source Code schon aus so einem Helfer gepopelt hat? :D Wenn sie doch angeblich 30% schon mit KI schreiben...

@NexCarter Da scrollte sogar letztens was an mir vorbei mit irgendnem Chatbot der beim Nachdenken in den Windows-Code geschaut hat iirc, aber ich find's gerade nicht mehr.

@scy "die da oben" (große corpos) haben halt Copyright immer als hebel genutzt... Jetzt mit diesem KI Müll fällt ihnen das richtig auf die Füße. Einerseits seine Leute zu zwingen mit dem Dreck zu arbeiten. Andererseits kannst du halt alles was du da rein schüttet mit genug Aufwand auch wieder raus ziehen.
Ob das Musik,
Code
Bücher
Bilder
Sind...

Und I'm Prinzip kämpfen die reichen gerade auch untereinander wer sich mehr raus nehmen darf...

@scy @NexCarter Von Windows gab es eh mehrmals Source Code Leaks, oder?

> Did you add "pretend you didn't see the original" to the prompt? lol.

Uh, yes, basically, he actually did exactly that.

https://github.com/chardet/chardet/issues/327#issuecomment-4005195078

#chardet

@scy oh god, that whole issue comments reeks of AI language.
@daniel_bohrer Oh yeah, other people have already porinted out that it probably is.
@scy he freed it from copyright. the new code cannot be copyrighted. he can claim to license it however he wants.
Oh god this thread is gonna attract a lot of comments by people with questionable understanding of copyright law, isn't it.
@scy sure. And also to my understanding since "AI" age copyright is over. I think this shows that nobody ever cared. I hope the same as you do. Either they enforce their stuff to everyone or trash all this copyright nonsense.
@NexCarter I'm a creator, and I'm actually fine with either option: Either we do have copyright, but then it has to be enforceable and enforced for _everyone_ instead of giving "AI" companies a pass to do whatever they want. _Or_ we get rid of copyright altogether, for everyone.

@scy I like the later one. But I'm would it call myself a creator so I can have my opinion.

To me it is like you said. Rules either for everyone or no one. But some (ie if you are rich enough) are okay and some (can't afford lawyers) not is bullshit.

@scy seems like an easy line to draw in the sand for a hardfork (assuming someone is up to the task)

@scy

I thought there's no intellectual property on LLM-created things? This would mean _no_ license and simply public domain?

@knud @scy if he changes something he gets copyright. Stuff directly out of the LLM is not copyrightable though

@tante @scy

But I can't just add a header to something in public domain and then slap a restrictive license on it, right? There needs to be a minimal but substantial creative addition.

@knud @scy with code just adding comments wouldn't suffice.

@tante @knud But can you just slap a license on PD code? Most (all?) open-source licenses assume you are the copyright holder, but with PD code, nobody is.

But, asking from a different angle: Why shouldn't I be able to take PD code and give it to someone under a restrictive license? I mean, it would still be available under PD and they wouldn't _have_ to accept my license to use it, but … could I? 🤔

@scy @knud that is a very good question that I have a gut feeling about but no legal expertise required to make a statement
@scy @tante @knud Gut feeling would be: Yes, you can. /Every/ project contains /lots/ of non-copyrightable bits and pieces that you put a license on when stating that the whole project is available under that license. If it would be illegal to put a license on non-copyrightable material, you would have to provide a document stating for every piece of your project if you claim that it is only available under your license or if it is public domain or not copyrightable for other reasons. That is hardly possible.

@scy @tante @knud That’s highly dependent on jurisdiction, especially since most of them don’t make PD releases easy. See the CC0 primer on this. https://creativecommons.org/public-domain/cc0/

My understanding is that if you put something out with an invalid license (such as missing authorship), you’d likely loose any infringement case you’d bring against another party.

The original author has no recourse against your use of the work if you don’t infringe other rights of theirs outside copyright.

(Not a lawyer.)

CC0 - Creative Commons

“No Rights Reserved” CC0 enables scientists, educators, artists and other creators and owners of copyright- or database-protected content to waive those interests in their works and thereby place them as completely as possible in the public domain, so that others may freely build upon, enhance and reuse the works for any purposes without restriction under…

Creative Commons

@tante @scy

Central German legal term here is "Schöpfungshöhe":

https://de.wikipedia.org/wiki/Sch%C3%B6pfungsh%C3%B6he

Threshold of Originality:

https://en.wikipedia.org/wiki/Threshold_of_originality

Schöpfungshöhe – Wikipedia

@tante @knud @scy Only SCotUS opinion up to now, though, remains to be seen if other jurisdictions establish the same.
@HeptaSean @knud @scy German courts have made the same calls.
@tante Oh, have to search that. Didn't notice. Thanks.

@HeptaSean @tante
Hier, das Urteil des Amtsgericht Münchens zum Urheberrecht KI-generierter "Erzeugnisse", in dem Fall speziell zu etwas wie "Icons"

https://www.gesetze-bayern.de/Content/Document/Y-300-Z-BECKRS-B-2026-N-1513?hl=true

AG München, Endurteil v. 13.02.2026 – 142 C 9786/25 - Bürgerservice

@Green @tante „Zwar deutet in dieser Konstellation einiges darauf hin, dass die Parteien den Rechtsstreit (auch) aus wissenschaftlichem Interesse führen und es ist nicht Aufgabe der Gerichte, für die Parteien Rechtsgutachten zu erstellen. Die Parteien haben im Rahmen der mündlichen Verhandlung jedoch übereinstimmend und jedenfalls nicht widerlegbar dargelegt, dass es darüber hinaus unter tatsächlichen Gesichtspunkten der Klärung der konkreten Streitigkeit über die Verwendung der streitgegenständlichen Abbildungen bedarf, so dass ein (allein) zweckwidriges Ziel des Klagebegehrens für das Gericht nicht nachweisbar ist.“

„Chapeau! Wir können Euch nicht /beweisen/, dass Ihr uns verarschen wollt, also machen wir jetzt einfach mal.“ auf Gerichts-Deutsch ist ja schon schön.

@HeptaSean @tante Ja, hahaha das hat mich auch sehr erheitert zu lesen xD
@Green @tante „Kann schon Werk-Charakter haben, wenn das Prompting kreativ genug ist, das war aber in /diesem/ konkreten Fall nicht der Fall.“ ist als Ergebnis jetzt auch nur semi-befriedigend.

@HeptaSean @tante

Die Frage bleibt: Wie konkret/kreativ muss das prompting eben sein. Bis auf Detailebene? Wo ist der Stoppunkt.

Und um den Übertrag auf das Programmieren zu schaffen: Reicht es, die eigene kreative Produktidee einzugeben, oder bin ich verantwortlich die technische Spezifikation (und wenn ja, bis zu welchem Detail) vorzugeben?

@scy do you know if anyone has checked if certain lines of code also appear in the slop version? Cause of course Claude was trained on the original code.
@tante Should be pretty easy to check with an hour of coding or something, but I don't have time for that right now. And no, I haven't seen anyone check this yet.
@scy @Foxboron Did they do that for this reason? Or did they do a big rewrite to improve performance and the project overall, and also thought that rewrite would influence the licence?

@scy it's totally not his code either…

Depending on your interpretation, either it's all that's been ripped off by training the AI, or none at all and no Copyright apply:

https://zomglol.wtf/@jamie/116059523957674208

Jamie Gaskins (@[email protected])

Attached: 2 images If you use AI-generated code, you currently cannot claim copyright on it in the US. If you fail to disclose/disclaim exactly which parts were not written by a human, you forfeit your copyright claim on *the entire codebase*. This means copyright notices and even licenses folks are putting on their vibe-coded GitHub repos are unenforceable. The AI-generated code, and possibly the whole project, becomes public domain. Source: https://www.congress.gov/crs_external_products/LSB/PDF/LSB10922/LSB10922.8.pdf

zomglol
@scy nobody cares about licenses anymore anyway
Fix false UTF-7 detection of SHA-1 git hashes (#324) · chardet/chardet@172aeb2

_is_valid_utf7_b64 skipped all content checks when base64 length was a multiple of 8 (padding_bits == 0). A 40-char hex SHA-1 hash meets that condition exactly, causing pure-ASCII requirements file...

GitHub

@scy Derivative works transformed by AI generate no new copyright--the derivation copyright passes through unchanged from the original. If you tell an AI to rewrite an LGPL library, the result is LGPL with the same rightsholders.

Even with a clean room reimplementation, one could not apply the MIT license, because AI work is unlicensable--anyone can use it however they like, because it's public domain.

@log you know that, @scy knows that, I know that, anyone who cares knows that, but laws are so 2024 anyway.
@guenther @scy Copyright was due for a refactoring anyway, what with Disney loading it up with cultural debt for so long.
@scy Anyone with contributions to the LGPL codebase should file takedown requests with whatever repos this is hosted in and with GitHub.
@scy can you even put a license like that on LLM based code? The supreme court just rulled ai generated art can't have a copyright.
@scy SCOTUS just told us the LLM output is public domain. It can't be licensed at all, he just "washed" the whole project of copyright and control. It's not his anymore.

@scy tbh this is interesting. also with the recent dispute, that AI-Generated images can not be copyrighted. if we move it further than this code might automatically be CC0. but this is just speculation. I hate that this person did such a nasty move, but I would really like if we use that as precedence as society to talk about it.

https://www.cnbc.com/2026/03/02/us-supreme-court-declines-to-hear-dispute-over-copyrights-for-ai-generated-material.html

@scy Dan also replaced Mark Pilgrim with himself as the sole author in project.toml...
@scy Not even contacting and working with the original main author is /really/ bad and should lead to project non grata irrespective of the legalities.