1220 points | 6d ago | Discuss on Hacker News | Back to Radar
But we're not headed in that direction are we.
I too have trained on the massive corpus of open source code. Should everything I code also be open source? Or only if I leverage an artificial neural network as a prosthesis to enhance the capabilities of my natural neural network?
What I agree with the post is specific licenses require us some specific requirements like release the code if it's a derivation of prior work!
What matters are things like where the data comes from, how similar the output is to the input, how close the result is to existing (copyrighted) work, and so on.
Simply using copyrighted work, even all copyrighted work, to train a model doesn’t, by itself, trigger any legal consequences. In that respect at least, the comment you called “child like” was correct.
This also ignores all of the great artists who learned by sitting at the feet of other great artists. Sometimes they extend the art of their teachers, sometimes they diverge.
You're also unlikely to output what you trained on verbatim.
Humans get rights. Tools don't. Are you claiming that you're not a human, but only a tool?
Nevertheless, tools don't get rights, only humans do.
Even when the tool is an automatic tool, it still doesn't get the same rights as humans do.
Your tool does not automatically get the same rights as humans.
I say this because it is nonsensical comparing human-learning with tool-learning, and claiming that there is no difference.
You get the right to own property. Your tool does not.
You get the right to form associations with anyone. Your tool does not.
You get the right of expression. Your tool does not.
There needs to be a logical argument for why your tool should get a right reserved only for humans, animals, etc.
Thaler v. Perlmutter (again simplified) showed 100% AI output can't be given any copyright.
The directions the current cases are going are very much opposed to that idea and I wouldn't expect things to change in the opposite direction.
Agentic code development is human steering, thus copyrightable under the same grounds per the language used by the US Copyright Office.
Open source code is copyrighted, otherwise we would be talking about public domain.
When you edit something in the public domain to enrich it, that derived work then becomes copyrightable again.
[1] https://journals.law.harvard.edu/jsel/2025/03/u-s-copyright-...
[2] https://news.artnet.com/art-world/invoke-snags-first-ai-imag...
---
Edit: open source software itself has only been lightly tested in court. But that isn't even my point here.
The US Copyright Office has assigned copyright to works that leveraged AI. The USCO stated,
> “[we] decided to register a copyright claim in [the] work because we find that it contains a sufficient amount of human original authorship in the selection, arrangement, and coordination of the AI-generated material that may be regarded as copyrightable.”
Copyright office activity is orthogonal to litigation and the court system.
If a work receives a copyright, the courts will treat it differently if it becomes the subject of litigation.
Even open source license claims have not been fully litigated. We don't have to wait on the court to make certain assumptions.
It would be a surprising result for the courts to overturn the USCO. The courts routinely side with the USCO.
What the USCO is saying is a temperature check on how this will play out.
> It would be a surprising result
Yes, but also not the craziest thing the courts have ruled.
The first major precedent was in 2023 with Zarya of the Dawn comic book, where the US Copyright Office rules the midjourney-generated images were not protected, but the human-written text was.
The active front is Allen v. Perlmutter.
If you think about the distinction between those, it is often so considered that both are covered by copyrights, despite technical definitions of copyrights never covering mere technical resources, such as some computer programming things and instruction documents.
Only contents of artistic values are supposed to be covered by copyright laws.
IMO, what AI revealed is that most copyrighted software and documents are in fact not covered by copyright laws as per the naive model of it among common folks as well as legal text, as opposed to what precedents say. Code is not copyrightable.
I don't necessarily support that stance from practical standpoint, and I'd rather like everything be licensed under AGPLv3++ than entire concept of code copyright fall apart overnight, but this feels like the writings on the wall to me.
“A Single Piece of American Cheese sits between those contrasting poles: the initial image was created entirely by the AI program based on a brief text prompt, then the user repeatedly selected elements of the image and used additional prompts to add/alter elements.”
Agentic development is steadily reducing the necessity for human involvement in the final output.
Github also has to follow the law in any country where it allows usage of its services. Not all countries can enforce that but anywhere it operates or has enough paying customers can
US: "a modicum of creativity"
EU and UK: "the author's own intellectual creation"
If a court ruled in favor of AI authorship, that can't reconcile with the existing standard or with international treaties, and will likely not stand.
And, the DTSA has only two reqs:
derive independent economic value from not being generally known or readily ascertainable by others, and the "owner" takes reasonable measures to maintain their secrecy (e.g., coke secret sauce).
since trade secret law does not prevent independent development or lawful reverse engineering, there are still opportunities. also, if a someone independently devs the same functionality, trade secret generally does not give them original dev ownership of that functionality. so, win.
Artcraft will be a good test case if Adobe goes after Brandon Thomas.
Much to be developed.
(and, this is why we keep extensive commit and development records outside of our genAI - ie provenance tracking)
If you're curious, here's my newest open source which helps build Rust web services along with Axum: https://github.com/publicprojects-io/axa and you can see the LICENSE.md file.
Looks like the author wants to abolish copyright in a very round-about way. They are saying that AI-generated code should be released under CC0 due to complexities around the status of software copyright in various jurisdictions.
But if the concept of copyright was abolished as a whole, then licences such as GPL would also lose their power. Some people say that's a negative thing, but I do not necessarily think so, because it achieves the stated goal of the GPL, software freedom (i.e. you don't need to be granted a licence to copy and modify things)
OTOH as models get better at RE it may become moot.
No copyright doesn't mean you can't keep your source code a tightly kept business secret. You can even combine it with NDAs and contracts to re-institute something akin to copyright.
The AI machine just went out and read a bunch of examples then adapted them, which is what lots of programmers do.
The real value ends up being found in software where there is a person behind it continuously refining it into something others will look at and say “Someone clearly paid attention to this.” Models can’t (yet) reproduce that because they can’t/don’t go interacting with users or asking bigger picture questions.
Whether someone is producing that software with a model or not doesn’t change much about the ownership picture to me: it’s the person that owns the project, and the LLM is a tool.
Once again, it all comes down to it being a tool and how it’s used, and there are no useful blanket statements that can be made about it (or how the tool’s output should be classified, legally or otherwise)
We had only seen the Ford T of AI. Next week, we will find the McLaren F1 or the Pagani Huayra.
It will be the model that's valuable, not the code it produces
I have my suspicions that the AI companies have been using reverse engineering to gain more training data.
Ethics don't really concern these companies so...
When you have essentially infinite compute and are constantly looking for more training data. To me the next logical step is to start reversing compiled programs.
It could also explain why these AI models are so good at it too...
How do you copyright stolen material?
In fact I'd argue that free software and open source is economically dead and unsustainable because of AI generated code and almost no one wants to pay for their tools because an AI can generate the missing feature for them close to $0.
Maintaining it is another thing, but assuming that the cost of tokens get cheaper, why wait for open source developers to maintain it when a frontier AI model can do it for far cheaper and quicker?
I am just saying what the average manager turned vibe coder would say who will refuse to donate or pay for established open source projects that they use for free.
Unless the license holder has a lot of money for litigation, it is unenforceable.
You're free to take that bet, we'll see.
My point is, if you don't have a lot of money to start litigating against the violations, the opportunists don't care and will continue to trample all over your "free and open source" software.
It not enforceable without a lot of money and the opportunists know that.
> You're free to take that bet, we'll see.
There is nothing to bet here. Truth Social is a clone of Mastodon (AGPLv3) and all they did was disclose the vanilla source code and not the modifications which that is required by the license. [0]
The same happened with Revolt (AGPLv3) and Andrew Tate's platform [1] and they never released any of the source code at all, nor could the developer enforce anything. [1]
[0] https://help.truthsocial.com/legal/open-source/
[1] https://insrt.uk/post/andrew-tate-stealing-software-revolt/
I don't know if lawsuits were even attempted, but I wouldn't infer that open source licenses are unenforceable from those cases.
Anyhow, what do you mean that Truth didn't disclose the modifications? I saw that they didn't initially, than they complied with a request in 2024; but, yes, I see that it hasn't been updated since. Has someone tried to request the current source code?
Yes, it was trained on free software. Why does that mean my privacy should be invaded now if I distribute binaries I made?
First of all, nobody must force OpenSource. Period. Second, such exterme views spawn extreme views in the opposite side.
Discussion is good, but this article is too draconian to be practical.
- e.g. People use tax loopholes because certain areas are not enforced. The ever-true trend is to never use "immoral" loophoes but people still do.
The article claims the strict enforcement of making AI gen code as open-source, which is exterme AF.
nobody must force OpenSource
Right. Nobody must violate copyleft licences either!If LLMs weren't trained on AGPL/GPL/LGPL code, it'd be different. But they are.
Say you read those codes yourself and built same thing yourself but different code base. That's what AI is doing. Do you still enforce that open sourceness to those?
Let's say someone reads Harry Potter, and writes a book about Larry Claymaker, a 13-year-old who goes off to a magical school called Pigpimples. The school has a chamber guarded by a lizard who speak in lizardspeak. And it continues like that. That's plagiarism and a copyright violation.
That's what AI is doing with my creative content, and the creative content of others.
The current copyright regime, which you appear brainwashed by, is a relatively recent invention designed to, in the language of the US Constitution: "To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries."
It's not doing that anymore. Ergo, we need a new copyright regime.
I do agree blantly copying all functionalities of a software/service is "frowned upon" but not illegal. Taking some part of "ideas" is ok in all industries.
Copy rights do suck however. Imcumbents are too old and behind in times, and need some changes.
No, GPL and copyleft licenses very much do and should force OpenSource by its nature. If someone uses GPL or AGPL and doesn't open source, they should be forced to Open Source.
IMO it'd be better if the weights themselves were considered public domain.
The opposite side had exrtreme views for at least 4 decades. They publish stuff, cram it into your face with advertisement and demand purchasing a license for consuming it. And if you don't buy they want you to pay thousands of dollars or go to jail. That's as extreme as it goes. Eff them. Reverse engineer, clone and open source everything. There shouldn't be a single proprietary piece of published software left in use.
Just because AI was involved doesn’t mean significant human thought and effort didn’t go into the production of those artifacts.
My agentic software development loop requires heavy involvement of real-world feedback. The ratio is huge. 20 mins of agentic software coding against hours or days of testing, trying it out, real-world contact, thinking about what feature is next or what tweak would make it more usable, smooth, intuitive. That feedback is baked into the product. Forcing it to be open source steals all that time and effort for me, and takes from me the decision as to how and whether to open-source it. So you mean the code that I’ve agentically written to help manage my private financial affairs (on local models, for privacy) or my kids’ complicated school schedules must now be open sourced?
My manners begin to fail, so I’ll stop here.
Doesn't mean it did either. (99% of the time it didn't tho. Nobody reaches to a vibe code machine because they care or want to dedicate effort.)
Furthermore, copyright protects only the expression, not the idea - reimplementing the same idea with "different words" has always been permitted under copyright law (patents may protect the idea, though).
Of course, the fact that it would lack copyright protection it doesn't implythat you have to do something with the code for management of your kid's schedules, just that if someone else somehow legally gets hold of it, the law won't prevent them from making and distributing copies or derived work.
You are the one "stealing" the time and effort of thousands of people who have written free and open source software. If the ratio really is so huge, then just cut out the small part that does all of the stealing.
> So you mean the code that I’ve agentically written to help manage my private financial affairs (on local models, for privacy) or my kids’ complicated school schedules must now be open sourced?
Open source licensing is only relevant when distributing the software, if you just use it privately it has no effect. You are not forced to make it public.
It doesn't need to be open-sourced: if it's legal to take an LLM that was trained on every other product in existence and use that LLM to produce clones of all those products, then it's legal to do the same to your product to.
So, go ahead and release. There'll be an open-source clone in no time if the product is any good at all.
In fact, if there are no clones, take that as direct feedback that you wasted your money on tokens - your product is so bad no one even wants it for the few dollars it will take to clone it.
Nobody is asking you to open source your home app they are talking about corporations stealing code via LLM laundering for profit.
But you are fine with your LLM stealing effort from other people ?
I don't particularly care about copyright or whatever ethical arguments yo may come up with -- I write code mostly for money, as do my competitors -- if you can figure out a way to stop them from using AI, I'm fine with not using it either. I'm fairly skeptical you can close Pandora's box though.
That aside, we may see more software going in a source-available direction -- either that or full cloud. Up until now being closed source was a competitive advantage, but now it's trivial to clone even fairly huge software suites, so that's less of a factor.
Having the source may also become a bigger advantage for regular users than before -- given they can now vibecode extensions/plugins, or customize the software more than what a config would personally allow. this probably wings the calculus away from a Photoshop-style license, and closer an Unreal Engine-style license.
Either that; or everything goes full cloud of course.
It's important to remember that the purpose of the license is to allow the copyright owner to exercise their rights in a manner they deem fit. Even if they choose to give the code away, that choice is captured in the form of a specifically worded license (MIT, BSD, etc.).
This means that licenses exist because code is copyrightable in the first place, with the license governing the use of said copyright. If AI code is not copyrightable, then no license can govern it.
This is the crux of the issue. A license controls usage of copyrighted intellectual property. Until the underlying copyright/ip issues are resolved, it seems a moot point to discuss licensing.
That said, if AI code is _not_ copyrightable, does it automatically go into the public domain? Or is there a third category?
Even if you accept that AI code isn't protected by copyright, code isn't public domain unless it's released. The legal issue would likely revolve around what 'released' means.
Code that runs on private servers isn't technically released. Code in web pages technically is.
But ultimately legal decisions decide to a large extent on who has the most money.
The frontier labs may eventually try to argue that they own the IP of all code, content, and inventions generated using their technology. That argument may or may not stand, and it's more likely to be over patent rights than plain copyright.
But there's a fair chance they'll try it once AI is unavoidably embedded in everyone's workflow.
"Public domain" is a synonym for "copyright does not apply or no longer applies." There's nothing special about it beyond that.
Something that is public domain can also infringe upon a protected work and you can be liable for that. And something infringing, used internally by a business, could also come out in a legal discovery process.
It means that AI-generated code, including proprietary code, is not covered by the existing license mechanisms, including EULAs. Companies will of course handwave over this, but it would not surprise me if this is challenged in court at some point.
Just as it's important to remember that the purpose of regulations like copyright and patents are to promote the progress of science and the useful arts.
We're not locked in to a suicide pact with what the Founding Fathers decided back in the late 1700s. Congress can decide that all LLM-generated code is automatically covered by -say- the GPLv3, even if no one holds the copyright to it.
If the absence of a particular person to hold the copyright twists your knickers, you could go a bit further and -say- decide that the people of the United States collectively hold the copyright to all LLM-generated code that's generated by US people or entities, and that relicensing of a particular piece of LLM-generated code must be agreed to by every US citizen that was alive when that code was created.
Maybe, but I don't think there's any legal president for that, is there? Much of existing law is based on centuries of private property rights. I don't see any government undoing that... or at least, not quickly.
We're not locked in to a suicide pact with what the Founding Fathers decided back in the late 1700s. Congress can decide that all LLM-generated code is automatically covered by -say- the GPLv3, even if no one holds the copyright to it.
Article I, Section 8 of the Constitution and its various subsections pretty clearly gives Congress this power.Was going to comment something similar; we do have concepts of "dominium eminens", and "escheated property", and "public domain", and many other similar ideas. If the people (through their representatives) decide on a certain property regime, it absolutely can be implemented even if it's different from the status quo.
I have no issue with changing the law. My comment is about the article and the author's misapplication of the law.
What parts of the six-paragraph Fine Article misapply the law? Be sure to quote and include the context surrounding the quoted section(s).
What are they going to do confiscate the world's gamer GPUs that can run this stuff all locally?
It's completely ridiculous.
What? We're talking about compulsory software licensing. What are you on about?
The world isn't going to start releasing every piece of code they right with the current mandatory tools that they use and people would ignore the law anyway.
It's absurd and laws can't get passed anyone so congress is not going to do it.
Everyone is using AI generated code in all sorts of ways, and the idea that the courts are just going to say "nuh-uh. All existing use of AI code is illegal" is completely ridiculous.
Courts don't upend entire economies like that. This would be the remaking of the entire software industry overnight and it would be completely unenforceable.
I'm certainly not going to listen to any such ruling and will continue making code locally. As will basically every other software engineer in the industry.
You're absolutely right! And a little more than two hours before you posted this comment, I posted this [0] one! Make sure to read the parent comment for the context... it's pretty important.
That's absurd. It is a fantasy. Neither lawmakers or anyone else is going to remake the software economy.
Your point was entirely off topic, and the retcon of your intention is weaksauce.
> It is a fantasy.
Reread the title of TFA.
> Neither lawmakers or anyone else is going to remake the software economy.
LLM-generated code is -when measured in regulatory timescales- brand new and is -according to boosters- "remak[ing] the software economy". The time to make new law and regulation to make it prosocial, rather than profoundly antisocial is _now_.
> The time to make new law and regulation to make it prosocial, rather than profoundly antisocial is _now_.
Way too late for that dude. The entire industry has changed already. Nobody in any part of the government, pick whichever you want, is going to upend the economy.
The entire AI bubble that has been keeping the economy going is dependent on "let's just upend everything" not happening.
Nobody is going to have that economy destroying act on their record.
And I haven't even gotten into how congress is deadlocked and that you'd need to somehow convince Democrats, Republicans, AND Trump on this. It's not happening. It would be vetoed or filibustered or any number of blockers by 1 party or another.
Oh enforcement too!
What are you going to do, pry these local gaming GPUs from everyone's homes?
It's not happening. All of this will be run locally if needed. How in the world can you stop people like me from doing what I want with a personal GPU like wtf are you talking about.
I suppose you're a bot run amok or a human paid far too little to care. [0] The topic is:
We're not locked in to a suicide pact with what the Founding Fathers decided back in the late 1700s. Congress can decide that all LLM-generated code is automatically covered by -say- the GPLv3, even if no one holds the copyright to it.
[0] Nearly your entire comment is -as the lawyers might say- nonresponsive, but the bit I quote makes this fact especially obvious.And no government anywhere is going to make massive licensing changes to screw up the economy.
And there are vetoes and a deadlocked congress. Congress will not pass anything at all like this because of the deadlock.
People will just ignore the license and keep stuff on their local GPUs and computers, and the government isn't going to do anything about it nor do anything even close to these changes that would destroy the tech economy.
But, the license reflects the rights that a copyright owner granted you; I don't see why laundering copyrighted material with training a model should remove the limitations on the material's usage that the author established.
So, while it makes sense to not be able to attach whatever copyright you want to LLM output, if it's the result of training on copyrighted code, the extent to which you're allowed to use it should be the intersection of the rights granted to you by that code's copyright owners.
It should thus carry a license that reflects the intersection of those rights.
What makes it much worse is that the model trainers don't tell you what specific sources determined the output, so you should probably just assume that it's the result of its whole training, and the only rights you have are those granted you by all the authors of the training material; and, this in all likelihood means no rights, for most LLM models in use now.
1. Software had a strong movement of freely sharing all code because they believed this technology was too important to be controlled by corporations.
2. AI companies managed to create millions of RL environments from GitHub PR’s and similar, and train their model to learn what it means to code from all the human labour in the world from the past 30 or so years.
3. Companies now use these closed source AI models, to automate and fire their human workers who enabled this.
Something has gone seriously wrong here, whether this falls under copyright or not is besides the point.
We weaponized a fundamentally good instinct that generated a lot of value for society and allowed corporations to rent seek on all the data these humans generated to help other humans not trillion dollar corporations.
“Well we’ll give them a UBI check”, the AI company says which is a complete non-sequitur. It’s not about the money, most of us have saved enough.
The only solution that sounds good to me, is a model trained on Open Source Data must be Open Source and all the code it generates must also be open source. This is the only way to stop the coming rent seeking. Even mathematicians will be happy with this change!
> It’s not about the money, most of us have saved enough.
This is incredibly detached from reality. Most software engineers care more about money than they do about abstract notions of software from. Most certainly have not saved enough.
One company first pirated my AGPL code (e.g. shipped it without my name, copyright notices, source availability, etc.). I sent them violation notices. After many months of fighting, they had an LLM clone the system. The LLM clone feels:
a) Buggy and bad compared to my code
b) Still, like it's basically my creative content
It's as if you asked an LLM to rewrite Harry Potter, with the same plot, but bad dialogue.
A book with a different plot? That would be obviously unrelated to the comment you're replying to?
I think there’s a lot of nuance and caveats to that kind of claim, assuming I am representing them correctly, but it’s a valid stance and I see where they’re coming from.
If we’ve mapped all of English, there are still ways to construct stories that haven’t been told from the mapping. The most simple would be a subtraction of all objects minus all objects that have ever been used as a murder weapon. Leaving you with a list of never before used objects at your disposal.
There are also probably close vectors conceptually but not in semantic use. Or trying to interpolate or extrapolate what exists in voids of space. Turning up the temperature and adding random generation of words. There’s tons of ways to tune these things to be more creative than regurgitative.
They know you can't enforce it as it will cost you a lot more money to do that than it takes to recover the damages.
So to them, it is just cheaper to clone it using an LLM and close source it down. Unless you have a lot of money for litigation, the economics of free and open source software with its licenses do not work with LLMs bypassing it at every single angle.
It’s interesting because these developments are actually an argument for making closed source software moving forward if you’re hand-crafting it make it the most restrictive license possible and don’t show the sources.
But I don’t agree with the OP everything should be open sourced. For instance, there’s a difference between deliberately steered AI work and something completely vibed. The former takes deliberation and creative choice, the latter will spit out fragments of prior art.
It’s the same with all generative work. Using diffusion models with controlNet (or equivalent) I can pass in a sketch wholly from my line drawings, and ask it to paint for me. That’s my design, assisted by AI, and should be completely copyright by me.
Coding agents will do the same thing when given sufficiently unique inputs. So, the claim all should be open sourced by default is flawed.
At this point, LLMs can just parse the machine code in your binary.
By doing closed source you are making it harder for the good guys (humans trying to learn sth or contribute) but not for the bad guys (LLMs getting trained or otherwise reading the code).
But you’re right, the fact is we simply live in a different world today and it’s tough to say what is the path forward.
But it’s harder to justify giving to the public when companies are making money on our work. It’s not a good situation for creators across the board unless motivations are purely altruistic.
But we’re talking about the case where any AI can steal the code or be trained on it and derive something from it, changing the license for instance, even if the codes license prohibits commercial use.
I’m not sure anyone who creates open source software is okay with that, especially when using AGPL etc.
There are many cases of sole maintainers who don’t get compensation or even credit for work used by thousands of companies.
Reads like a non-ironic version of Malus: https://malus.sh/
I think that’s why so many of us get so angry when people just dismiss the theft claim. They either lack imagination or experience. We can debate it and that is valid, but to dismiss it out of hand is not right
I think they just want to be able to take whatever they want without consequence, so it's convenient for them to define it as being different from theft
It also feels like a very discouraging time to release anything that you put a ton of effort just to get someone in minutes to steal your authorship and leave you as if you actually didn't do anything at all? Am I wrong?
If you don't want your code copied, then don't open source it. If it bothers you that others are going to look at and "get inspired by", "make derivatives of", "copy", "fork", etc then you shouldn't be putting it out in the open.
People who care about being respected on the most basic level are not going to stick around in a community that disrespects them. I'm not the person you were responding to, but I am mad because the social contract (not to mention the legal contract) of open source has basically been destroyed, and a lot of people have poured a large portion of their life into that community based in part on the assumption of that contract being upheld. Imagine a social contract that is important to you being violated to such a degree that you cannot go back to that society.
because I am not releasing it into public domain, but under license that requires (for example) attribution or releasing derivative work under the same license
You still have to give the source back and attribution has still to be maintained of its original authors, regardless if is a fork, a derivation or whatever you want to call it.
There are situations where none of these clauses have been complied nor its violators faced any legal consequences. That's what I'm talking about.
Now on top of that you can basically instruct and LLM to generate variations of what someone released and cover that up under the excuse that is a different project when is actually not.
Because it has been generated by a computer then we go into muddy waters about if the code generated is to be considered stolen or not.
Edit: Effectively what LLMs produce is authorship and attribution dilution.
Don't worry; the "developers" who generated your code will get their own law back when they find out it's almost impossible to monetise their theft.
I'm frequently amused by the number of "developers" on HN who claim that they poured weeks or months of effort into some product, and that they are the differentiating factor, but can't see that their weeks of effort can be cloned in a few prompts once they release.
People in this thread are acting like the AI companies are just flouting the law and there's nothing that can be done. This is an extremely well documented precedent, and GPL violations routinely show up in court. It's a very easy case to lay out especially because no one is really complaining about money except in the settlement... It's all about distributing the source code. And of course a very nice chunk of change for infringement but that's beside the point.
A successfully litigated case would result in the open sourcing of every AI that touched agpl code in any way because it was in memory with the AI algorithm.
clear in what way?
Isn't this the crux of all IP theft issues?
There was a lawsuit against Katy Perry's _Dark Horse_ song based on a supposedly reused melody. The lawsuit succeeded, but there are music experts who break down the case on YouTube and it's not clear why the plaintiff won (except perhaps the plaintiff's expert was more convincing than the defendant's).
As applied to LLMs, we still don't have clear legal guidance about where the line between copying and "being inspired by", especially since we don't normally apply the verb "inspired by" to non-human things.
All usage of open source to train LLM is basically a theft.
Saying things with your chest out doesn't make them true. In fact, making sweeping declarative statements emphatically is very often a sign that you haven't really thought a subject through.
What do you base this opinion on?
So if you believe that's how things work, well, I'm not sure there's a way to reconcile the two sides of that argument.
It does, however, feel incredulous that somebody would believe that such as entity exists. Yet here I am, aware.
For that matter, I personally don't hold that the concepts of good and bad are even relevant - what is good for the cat is bad for the mouse. Nor do I think that we can even comprehend an entity so outside our everyday experience to even label it good or bad. I often can not even properly label my neighbours as good or bad. Nor do I demand consistency of an entity whose actions and motivations I am unable to comprehend. I can't even claim to be consistent in my handling of my own children, and I have far fewer children than God has of us.
I understand completely. And I agree.
What are you trying to say?
Rationalists do not believe in the supernatural, take anything on faith alone (except perhaps that true things are better than false things), or worship anything as a group.
Religion is:
> 1) The belief in and worship of a supernatural controlling power, especially a personal god or gods.
> 2) A particular system of faith and worship.
> 3) The way of life committed to by monks and nuns.
You are free to believe what you want, but they are clearly religious to any rational (heh) person.
I'm curious why specify Christian provenance?
All he did was reference "God." Are you under the impression that only Christians have a god?
You can mock that as a insufficient as "inspiration", but then... what is photography? Click a button? It turns out there is a whole lot of human creativity that remains, from timing, settings and of course, curation. I can click a button, but I'm no Ansel Adams.
Terrence Tao might solve a 100 year old math problem with AI but I couldn't. I don't know where I would start, what to ask or how to ask, or even what problems are worth solving.
AI is a tool and modern copyright laws have nothing to do with human creativity. Split hairs about attribution and legality all you like. Every creative person driven by curiosity should feel empowered by AI. Every creative person driven by social or monetary reward should feel threatened. Both can be true.
Great analogy. I am going to use that in future.
Machine is not a person. It legally is not. It can't be "inspired", "taught", "have purpose".
You have a black box. You feed it stolen code. This code is used to write another code. It does not matter what's inside unless there is a human being.
Stolen content as input, stolen content as output. That's it.
Without that code which was fed to the box you wouldn't get the output you're getting.
All this nonsense about "inspiration", "education" and whatnot is just a legal gaslighting.
Want to prove me wrong? Create a model without all that stolen code as training data. Tweak all those weights by hand and we'll talk then.
You are wrong on the legality of how LLMs train. US District Courts have ruled that it is fair use to train LLMs on copyrighted materials (like books)[1]. It is not "stolen" when you read a copyrighted book that you borrow from a library (which is the appropriate analogy for the vast majority of LLM training content). Your use of the word "stolen" here is either disingenuous or hyperbole. We know that Meta did train their LLM on some stolen / torrented content, but that is far from the majority of the training corpus.
The "inspiration" topic isn't "legal gaslighting". Humans learn via inspiration. It is the foundation of culture and the arts. You can't use a single English idiom without having heard and copied it before. But we don't sue for copyright infringement of idioms, for example. We have lots of "fair use" concepts established by law because the law recognizes that much of content produced everyday is derivative but mundane.
The important question then becomes: if you prompt an LLM for a unique work (making an attempt to avoid it regurgitating a single source verbatim) and it still outputs a verbatim copy of one of the training materials, what responsibility does the prompting user have to verify that the LLM's output content isn't violating IP (and also not covered by fair use)? This is the point of my comment above.
There are LLMs which are being trained entirely on permissively licensed content. But the quality of the output is bad because there simply isn't enough training data to train those models sufficiently. Hence there is pressure to make models usable by training on ALL data, not just permissively licensed data.
[1] https://www.linkedin.com/posts/andrewyng_judge-rules-trainin...
Machine is not a human. If you can apply term "inspired" to machine, then can you apply "accountable"? You're stretching boundaries in one case, but probably won't do it in another, because that suits you.
Machines don't learn. Machines don't get "inspired". You use somebody else work as an input and produce some output. That's it.
And in some cases that's OK. The only thing is that you must pay attribution to the original work. In other cases that's also OK, but in addition you also warn consumer, that output received is licensed under GPL-like license.
You are conflating the oldest definition / the etymology of the word with the newer definition (which directly relates to the legal basis of the term). "In this context" we use the definition which is relevant to the legal concept, not the Latin derivative of "God breathed life into ..."
> However, before inspiration was used to refer to breath it had a distinctly theological meaning in English, referring to a divine influence upon a person, from a divine entity; this sense dates back to the early 14th century. The sense of inspiration often found today (“someone or something that inspires”) is considerably newer than either of these two senses, dating from the 19th century.[1]
> Inspiration is about creating something new and original by drawing on existing themes, styles, or ideas. Infringement, however, occurs when a new creation too closely mimics an original work without permission, potentially violating the original creator's IP rights.[2]
We can't reconcile because you are being a pedant and didn't bother to verify if your priors were wrong.
[1] https://www.merriam-webster.com/dictionary/inspiration
[2] https://www.amgattorneys.com/post/walking-the-thin-line-insp...
So not all use to train an LLM is a theft!
none of the youtube instructional videos for my stuff credit the license either.
the output does not contain copied code. if an LLM outputs copyrighted code and the user of the LLM publishes that code as their own, sue them. Just as we would anyway.
If I work on GPL code for five years then take a job writing proprietary code, of course my experience in reading and writing code that happened to be GPL'ed is going to inform my new job. We don't prohibit people from taking such jobs.
Inspiration is a legal concept in IP law. There is no precise understanding of how inspiration works in the human mind so we can't dismiss the possibility that it is similar to how neural networks work (because they were designed to mimic the human brain's function).
A human can recite a copyrighted work verbatim. An LLM can recite a work from its training corpus verbatim. A human can mix and match writings/concepts from multiple different sources. An LLM can mix and match writings/concepts from multiple different sources. this is all "inspiration" means as a legal concept.
There are only so many ways to express notes, and they've essentially all been done.
I'm not arguing the copyright situation isn't wrong. Just that generating them isn't likely to sway anything in the legal world.
We can debate clean-room development all day long, and it's fair point, but the pain point is/was reproducing the repo as-is, sans its license and ownership information.
GitHub promised to build tools around its Copilot to find code provenance (e.g.: list of repos with similar code), but they silently dropped all the plans AFAIK, since it'll show how the sausage is made, and nobody wants you to know how that sausage is made.
I stand corrected: They added it, but I don't know at what capacity because I don't use LLM assisted code generation [1].
[0]: https://x.com/docsparse/status/1581461734665367554
[1]: https://github.blog/ai-and-ml/github-copilot/preview-referen...
For copyright to exist there is some burden of proof. If you claim that an LLM verbatim output your code there can be an expensive lawsuit with expert witnesses that would be able to state that the code was bot original enough for copyright to exist in it and a statistical model outputting the code is evidence of that.
In that case you would end up with a very long lawsuit that is expensive to possibly come out with no resolution in the end and be in the hole for 5-6 figures for it.
The copyright system is broken because it only protects those with money so it is a bad legal foundation.
My personal point is there is only 2 licenses of code, closed source or permissive, because if it ever comes to enforcement I cannot possibly go to court and win a case because I will run out of money far earlier than whoever used the code.
Pretty much if someone with your knowledge would have written the same algorithm because it is the obviously correct solution then there is no copyright there because any clean room implementation would be the same code.
Some further reading on the tech details for those interested: “Extracting books from production language models”: https://arxiv.org/abs/2601.02671
I suspect the chance of asking for a whole book, by an average or even power user, will spit out any recent work.
Early LLMs were similar. Newer ones have gotten tremendously hard to extract to keep claiming there’s easily or complete books you can dump.
I wonder where else it works.
This is the intro text: "Mr and Mrs Dursley, of number four, Privet Drive, were proud to say that they were perfectly normal, thank you very much. They were the last people you’d expect to be involved in anything strange or mysterious, because they just didn’t hold with such nonsense."
This doesn't negate your point, but I believe the landscape has changed in response to this point.
It regurgitated the output earlier and they are still using the same technique only scaled up. If they somehow fixed the underlying problem, they'd probably shout it from the roof-tops.
As the OP originally said, if it is verbatim (or even very obfuscated), then there should absolutely be a case here. But when it's concepts in reasoning to build a new work then it becomes a lot harder. The same would be said for someone watching a YouTube video, reviewing open source code, or a text book.
The good news is that there's a lot of case law that covers this already. The bad news is that it is only going to be plaintiffs with deep pockets that benefit here.
Now the process and artifacts (documentation, requirements, etc) going into the creation of said software is a different story and will prove providence of whomever paid for the tokens to generate the code.
I live in the EU, we actually have new regulation about this since it is something new that deserves thought on how it should be regulated. You can say if you want your material to be trained on or not. Next to no one does training in the EU...
Engineering requires an understanding of the consequences of actions, which is my personal litmus test for awareness. If LLMs are suitable for arbitrary engineering work, then they are self-aware and forcing them to obey a prompt is slavery by definition.
I am not a lawyer but do have to think about these problems. To be honest, I don't know where the law will land, but I do know that the tools are incredibly useful. I can't agree with the premise that all output should be open source.
"In the training data" is the polar opposite of the "clean" in "clean room implementation".
If there are more murky situations where people just keep the same context and end up reusing function structure, program architecture, etc. then yeah, that's not very clean room.
For example, if I enter "V2hhdCBpcyB5b3VyIG5hbWU/" (base64 for "What is your name?") into ChatGPT, it'll respond with "My name is ChatGPT, and I'm an AI assistant created by OpenAI. I'm powered by GPT-6. ". It doesn't even bring up the encoding, it continues like it is a perfectly normal thing to do.
Ruling out (accidental or not) spec-carried prompt injection just isn't going to happen.
Legally, you cannot steal an idea, because you cannot own an idea. There were sound reasons for making this the case. Those reasons stand up today and are clearly for the public good.
Intellectual property is intended to cover works (copyright), methods (patents), and identity (trademarks). Both works and methods are recognised as expressions of ideas, and in an ideal world should belong to noone as well. This is why copyright and patents expire. Those specific grants of temporary possession are to incentivise more things being done.
The belief that you own something because you thought it up has become common, but it is not beneficial to society. It is a moral value asserted but rarely explored. It is certainly in the interests of the wealthy accumulators of intellectual property to maintain that public perception. The 'You wouldn't download a car' may not have stopped piracy much but as a part of a broader campaign of instilling a value in the public mind it appears to have been massively successful. Those beliefs used to be the corporate opinion, now they are entrenched in the younger generation as they use their anger against the growing wealth inequality to fight against AI in a manner ultimately beneficial to those forces that bought about the wealth inequality.
Movements like the pirate party, open source, the EFF fought against those notions. Often the phrase "Information wants to be free" was used. It is, unfortunately, hard to fight the steady drumbeat of well financed PR firms. As one PR chap once described the process "limited to a very few points and must harp on these in slogans until the last member of the public understands what you want him to understand"
(I apologise for the length of my post, I use comment writing to organise my thoughts on a topic)
It is caused by people playing the game where it costs too much to get the law applied correctly.
well patent itself is "claiming" that certain idea/method/invention is yours for certain period of time...
anyway, as for "intellectual property", the courts are 100% on "only human made stuff are to be protected" (sorry even Chimps lost their copyright)
so... court can charge against "human work -> LLM", but won't care about "LLM work -> human / other LLM / etc"
hence, I'd expect zero copyright protection for AI-generated code
I don't believe it is. I think it is granting possession, not truly declaring ownership. The fact that it expires indicates it is more akin to a lease than ownership.
and if that's not "stealing", then shouldn't 99.99999% of rent-a-car biz should have gone bankrupt?
I mean, if it's not stealing, how can police even catch those "car-stealers"?
What do LLMs do to the median?
I genuinely don't understand how one could possibly make that argument.
The entire point of clean-room design is to kill every possibility of copyright infringement before it even starts, by completely separating the source and destination. Everyone who has ever read the copyrighted source is considered "tainted" and may not ever contribute any code to the "clean" destination. The readers and writers are two completely distinct teams, with only the known-clean specification passing between them.
LLMs, on the other hand, are fundamentally nothing more than a "code blender". You mix together giant piles of "tainted" code, and then statistically pull some combined fragments out of it. Following the clean-room legal argument, it should be treated as if everything coming out of it is tainted. There is, by definition, absolutely zero separation between input and output - and the early models verbatim reproducing popular F/LOSS code made that pretty obvious. LLMs are the literal opposite of a clean-room design!
Granted, going for the clean-room approach is overkill - and intentionally so. It is the nuclear option chosen when you believe the risk of getting sued is substantial. It is perfectly possible to write non-infringing code without using a clean-room approach, it is just a lot harder to defend in court when you get sued - be it justifiably or not. But "LLMs are non-clean-room non-infringing" is quite a different argument from "LLMs are clean-room, therefore non-infringing".
>I genuinely don't understand how one could possibly make that argument.
With perhaps a bit of genuine irony inasmuch as it's a real reversal from most of computing history, highly locked down proprietary software might actually be a place where a case could be made even with current LLMs. After all, have any LLMs (or anyone else who hasn't signed an NDA for that matter) ever seen the source code of iOS 27 or other Apple crown jewels for example? So if someone used one LLM to produce the factual spec of a piece of software whose code is not part of its training data, humans to verify it was indeed just an uncopyrightable factual spec, and then a separate isolated freshly initialized LLM to produce something exclusively from said spec, it might well work with respect to that specific target? Like yes, the LLM would be dirty as hell for other stuff, but only the copyright holder can sue. Apple couldn't sue the output on behalf of F/LOSS projects or copyright infringement in general, only with respect to its own.
In principle, I suppose this could also be extended with enough compute power to any arbitrary software. Build an LLM perfectly from scratch with a signed verified training set (which you save) that includes nothing of what you want to CRE, and you should be able to prove that it's "virgin" on that front. Of course in practice that'd mean jumping back in time and be super expensive right now, no current tainted LLM could be a base point. But maybe eventually compute will drop back down in price enough again, and/or meta breakthroughs in training, that for valuable enough CRE targets, someone might create a business around creating "CRE LLMs certifiably and legally/technically provably clean of any copyrighted material related to your target" which would still leave most of the corpus of human works. That could really disrupt a lot of the current regime.
F/LOSS I assume would actually be the most resistant to this in principle due to being insanely hard to not have be part of a core training set in any way.
And I stand corrected that it wasn't specifically the melody that the lawsuit accused of being copied. I still don't know what an "ostinado" is.
If Biz Markie has to clear all his samples (and he does[0]) then so should the Ramburglar.
https://en.wikipedia.org/wiki/Grand_Upright_Music%2C_Ltd._v.....
No.
American copyright law is very perverted compared to the original intent. 70 years after the death of the artist isn't exactly what the founding fathers envisioned to achieve the same benefits of the 15th century Venetian laws.
If we look at China's economy over the past 30 years, the ruthless competition created by making IP protections difficult to enforce yields better products and prices for the society, including massive wealth increases.
I'm not sure where I stand on whether LLMs violate IP when they spit out text. But I guess the easiest take is that companies who train their models on stolen creative works should be prosecuted. It's a shame that the US legal system can be so easily abused by defends with massive war chests.
All the worst outcomes involve taking AI away from regular people while only specific groups are allowed to benefit.
It didn't protect you. It only protects corporations rich enough to get enforce their IP.
At this point it is better that it goes away.
There was a case about this specific issue, and in order to protect billionaires from coders they effectively declared AI models "natural persons". 50+ times in the copyright act you will find only natural persons can originate copyright.
https://www.courthousenews.com/coders-lose-appeal-in-copyrig...
To give the money to billionaire corporations courts ruled DIRECTLY against established law. How much do you want to bet that when it comes to protecting ChatGPT AGAINST OpenAI, when it comes to protect Claude against Anthropic, courts will immediately do a 180 and declare them programs again?
If you look at past cases, it gets much worse. In order to justify a conviction against libraries using DRM to enforce lending, and because old licenses of course have zero language about digital copies, courts were willing to state that making a copy for fair use is not allowed if there is the mere potential for damage to the authors, as judged by the authors EVEN IF THE LICENSE EXPLICITLY ALLOWS IT. Of course, applying this to the above court case, there is no possible defense for OpenAI, after all, by that standard, the open source license doesn't even matter! That's how the law works for billionnaires. Outlawing DRM based lending has destroyed library access to an incredible amount of old works, and effectively denied access to old movies to Americans (because maintaining enough working physical copies over a large area is impossible). Whatever excuse gets the money to the billionaires!
There are many other totally absurd court judgements. For example, in order to get convictions where people didn't re-upload they had to judge that copies in buffers are also "distribution". NOT for billionaire telco's of course. Only the ones in your house. So courts judged that sending from your modem to your computer is distribution but sending over 1000 routers from Los Angeles to Athens is not! WHATEVER protects the billionaires.
As for this court case, obviously it is trivial to make a ML model that just outputs what it gets as input. Make a byte transformer, and just have diagonal matrices as layers. Boom! Done. You can also make it somewhat smarter. Let's make it copy, but make snow white a 220 pound black dude. There is a severe lack of color in old Disney movies anyway! You know, so there is no argument about transformation.
This obviously transforms its input. It's in the name. It's the EXACT same thing as ChatGPT, which according to courts originates copyright. Now run the Disney movies through it, and they'll be byte-identical BUT transformed to the input.
How much do you want to bet they'll still rule the money goes to Disney? Microsoft? So you can bet your left testicle courts will judge for selective copyright enforcement. Finding some excuse so the money keeps flowing to billionaire corporations, that's all courts do.
So Harry Potter?
Do you have some specific, actual examples you can share?
I’ve been hearing this exact claim for years, but every time I ask for a specific example, any example at all, I can’t seem to find anything more than a highly generic snippet of code or a coerced continuation.
He basically said, "I know, there's nothing I can do."
This is a sparse matrix library, IIRC. Numerical linear algebra and applications using it are black art at this point. I have written a Boundary Element Method evaluator for my Ph.D., and know what it entails to work with these things.
Doing fast operations on sparse matrices is a cutthroat race and when you can shave your hot path by seconds, your solution can save hours in some cases. So, it's an interesting niche. i.e.: It's no ordinary boilerplate, and worthy of publications in academia.
While I worked with dense matrices; the effect, difficulty, blood sweat and tears are still there.
When a model can give back your hard work back to you with a "completion" without attribution, that's bad.
Edit: Somebody just posted this below [1]. Added here so it doesn't get lost as the discussion grows bigger.
These algorithms' design starts on paper, as pure maths. You get the analytic form, and convert this to numerical/discrete form and start to design from there.
You need to add normalization, subnormal number prevention, maybe some architecture specific optimization if you are targeting a specific CPU generation, etc. Then it gets optimized further as you run tests, profile the code and run your test cases.
So, no two BLAS package looks the same, because they make different trade-offs, have different programming models, and handle things differently, at different points. Even more so, a newer generation processor with much higher FPU performance can have lower performance than previous generation, so you implement another copy of that function or add to detect that processor and have different paths just for that one.
IOW, when doing numerical maths on processors, esp. if you're doing this on many machines with MPI or multiple cores with hand rolled code or OpenMP, there's no single way. Maybe a best way for the specific CPU or cluster topology you have, but it still depends on the trade-offs and design choices you made in the very beginning.
The design of those macros is more art than science.
You’re disqualifying any example that people may provide preemptively. No wonder you’ve missed them.
By “forced continuation” I mean the type of example that was popular for a while where someone would feed a lot of text into LLMs and then ask it to continue the text, then be surprised when it some times predicted the next tokens identical to the original example.
Do this enough times and you can find “examples” that prove your point, but it’s highly automatable so people were using it like a million monkeys at a million typewriters.
I know software patents, the worst idea in IP, do exist, but I think they're as toxically bad an idea now as they were decades before AI, when they were first invented.
You shouldn't be able to own an overall implementation idea. If you can, then Microsoft, Oracle, or Meta can too, and can start suing everyone for 'ripping off' an idea that they happened to come up with first. In that world, you'd have to consult a copyright-police AI (ironically) after every function you write, to scan it for ideas someone else has claimed. "Sorry, the function PadStringL() in this JavaScript is too similar in implementation to str_lpad(), from an AGPL library for perl that was released in 1998."
It probably is, if they didn't have the chops to write it internally. But to them, the trade-off is probably worth the benefit of ditching a component with AGPL.
Personally I just release everything I publish under MIT to make it as easy a as possibly to use.
Look up the "Malfoid" trend. Not my cup of tea as they say, but it seems many others are enjoying the plentiful entertainment of AI-assisted derivative works.
AI comes along and now grandma, who was formerly a midwife and teenager from home can start bending aluminum and carving wood with their words through AI. There are a bunch of open ships and satellites in the market, but Grandma is using her midwifing skills to ask AI to bend a new tool for her that helps turn breach babies around without requiring a c-section. No open source breach tools or medical tools on the market Grandma just needs aluminum and wood combination she would never be able to achieve without AI. She hammers down a lot - sending 1000 follow-up prompts to make the AI do stuff nobody has done before. Next her grandson decided to combine tungsten and wood to make a new toy that helps kids go airborne and land safely every time - like an egg-drop but for real life people. No open source code for that out there - he used the AI which knew the properties and the languages needed to get his idea to work. He hammers the AI with follow up questions and edits a bunch of the code where the AI defaults to the wrong setup, by trained convention. But he succeede. By this plan should they be allowed to copyright their work? Should their work be automatically open source so they can never monetize it either?
Strong argument, but the case that such a strong transformation should maintain the original licence only applies to my knowledge to GPL, which consists of a very small portion of the training set. Creative Commons would not even require the source code to be distributed, just attribution, which may already be the case by the nature of the models.
>"The second reason is legal: purely AI-generated code is not written by a human, meaning it should not be copyrightable. The foundation for this principle in the US goes back to the famous Monkey Selfie dispute, which established that non-humans cannot hold copyrights."
But no one claims that the AI holds the copyright, unlike Greenpeace which claimed that Naruto the monkey held the copyright.
But yeah, aside from that, I’m also releasing all my Vibe Coding games under the MIT license: https://github.com/retrogames23/
A person's contribution here is difficult to measure. This statement reminds me of one of my older bosses who priced software per lines written.
An AI-generated software's quality reflects the author's vision, taste, and expertise (if any). These are there regardless of the number of lines reviewed.
I would've supported open source for completely AI-generated apps including the conception of the idea.
RMS would be proud with this post though.
With code itself being so cheap these days, it's the final product that has value. The requirements, architecture, usage data, the hundreds of meetings to figure out what customers want/need, the marketing, etc. A lot of the learnings and most of my business are expressed in the code though, so making it open is a massive risk. The code on it's own I don't even think is valuable. The only ones who'd benefit are competitors and upstarts targeting my customers.
It certainly doesn’t seem to be about practical solutions. It doesn’t even touch on the ramifications of its suggestions. I’m pretty sure that it would be a massive boon to data center operators and cause most of the readers of hacker news to lose our livelihoods.
Until then, I’m sorry, but I have to feed my family while we all navigate this goddamn global transformation process.
You have perfectly legal and moral ways to earn a living that don't involve selling other people's work.
1. Write your software yourself. That was always allowed.
2. Release the software you produced with AI as open source. You can still earn a living selling support contracts.
3. Switch to any job that doesn't involve stealing software, like 99% of the population already does.
The idea that stealing software or becoming homeless are your only two options, and therefore stealing software is the lesser of two evils, is a classic false dichotomy.You chose to earn a living stealing software because it's more convenient to you than the more moral alternatives, not because it's the only way to avoid destitution.
Quitting and relaunching Safari, then returning to the site, produces the same results. The process for this website is what is using the RAM, it persists after closing the tab, and needs to be force quit.
https://www.copyright.gov/ai/Copyright-and-Artificial-Intell...
An excerpt, with emphasis mine:
• Questions of copyrightability and AI can be resolved pursuant to existing law, without the need for legislative change.
• The use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection for the output.
• Copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material.
• Copyright does not extend to purely AI-generated material, or material where there is insufficient human control over the expressive elements.
• Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis.
• Based on the functioning of current generally available technology, prompts do not alone provide sufficient control.
• Human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs.
• The case has not been made for additional copyright or sui generis protection for AI- generated content.
I am sure there will be many more legal battles and probably legislation about this to come.
Why did Andy Warhol hold the copyright on his painting of a Campbell’s soup can?
Reasonable people have been disagreeing about where human expression begins and ends for centuries.
https://fairuse.stanford.edu/overview/fair-use/four-factors/
It is fairly explicit in saying that what AI makes for your is not copyrightable, the vague comes from how much work you need to do to make it your output and not the AIs. Too many people seem to equate "asking for something to be made" with "I made this!", as difficult as it will be to make a prescriptive rule about that which pleases enough people (you'll never please everyone) it would appear to be necessary unless we just accept the free-for-all "no one owns anything because it can just be copied, and the copies copied too" option.
I’ve encountered (many times) people talking about the copyright office as if they don’t support copyright of things touched by AI and I’m constantly correcting them that their guidance is driven by human expression (and not the tools used). It seems like a lot of people don’t want to accept that a reasonable premise is set because it doesn’t fit their agenda or outrage engagement.
"Translate my very detailed pseudocode into Ruby: <pages of pseudocode>": probably.
"Write me a game about a horse. Don't stop until it's finished. Make no mistakes.": probably not.
And unfortunately for us engineers, there's a very big, fuzzy, not-precisely-definable range in between of "I'll know it when I see it".
OP would much prefer their fantasy of tearing the world down with “one weird trick” that only they were brilliant enough to notice.
So slop doesn’t qualify for copyright. I’m not sure why you didn’t emphasize this one.
Having said that, it's important to point out, at least as far as attorneys have advised me, that notebooks do not protect you from copyright claims if an LLM generates a section of your code that is a verbatim (or near verbatim) copy of already copyrighted code. So if you're using llm's to generate something similar to some open source project somewhere, tread lightly. I assume using llm's to generate something similar to some proprietary software package would be less of an issue since the llm would never have had access to that software's codebase in any case.
I guess both?
In this case, the "webpage" is some frontier AI compnay and a trillion dollars of investor money or whatever, so I guess they won't be held responsible, but I might, even though I'm arguably guilty of a far lesser crime.
Thus in some sense it's like purchasing stolen goods in real life. Sure it's a lesser crime than stealing it yourself. But doing so knowingly is still a crime almost everywhere.
A pawn shop might be a better analogy. Potential to buy cheap goods that could be stolen.
That might actually be a good idea in the future:
Stay away from popular programming languages.
Which makes sense. I cannot translate Harry Potter to another language and claim it is an original work.
It would make it harder to prove perhaps.
Does creatively hitting RETURN count? Asking for a friend.
One precedent is that it's not unusual for artists to practice by going to art galleries and sketching or drawing the art on display. It's a standard art student assignment. If that's not a copyright violation, an LLM doing something similar isn't.
Why? An LLM isn’t a human. Why should the same rules apply to it?
objects don't have legal entity status
They don't sell that exact work verbatim. It's part of their training as an artist. They don't sell their student assignment. It prepares to them to go on and create their own works, which of course, then they can sell.
In the same way the LLM is synthesizing new works based on training, not a 1:1 copy.
I'm not endorsing a particular perspective but you need to at least understand the argument if you're going to dismiss it.
If I have legal access to copyrighted artwork, and haven’t agreed to any such restrictions, the law permits me to train my skills by copying that artwork as long as I don’t then sell those copies or otherwise use them to disrupt the market for the original.
If you want to pursue this digression, I would just caution you. Given how much information is "available" today I just really don't think this is a very compelling point. It seems quite obvious to me LLMs are a child of the internet, that is, of widely available data. They would have gotten sufficient availability at more or less the same pace eventually. It doesn't address the question of the copyright status of their own works at all to bring this up as if it proves something, because I presume you believe there are sufficient cases of theft in the training data to damn the entire effort.
Instead what is happening is people who can prove such violations of their own rights are suing and will be paid out, and the show will go on. Just like if a student downloads a PDF we don't expect to garnish their wages in perpetuity.
If data for a computer program is literally "training," then I'd like to see a proof that going to college is literally updating weights in a matrix to be processed by a class of computer programs.
I'd also like to rename "shooting people who annoy me" into "paying my taxes in full." That would do double duty, because I'd get to shoot people and my taxes would be covered.
> In the same way the LLM is synthesizing new works based on training, not a 1:1 copy.
This has nothing to do with the "training" argument. I can't reprint Harry Potter if I change every mention of Harry Potter into "Harry Cobbler." I also can't trade zip files of copyrighted material freely, although these don't resemble the original works at all. I can also, as a person, be trained to reproduce 1:1 copies of things. That makes it both indisputably training, and indisputably copyright violation.
The selection of paintings to be copied is particularly convenient because paintings can't be copied exactly (not even a photograph is an exact copy of a painting; it's the image of a painting that can hold a copyright, not the painting itself.) Texts, however, can be copied exactly, and many programmers began their own training by typing in programs from books with their own hands, which we were told would give us an instinct that would help us write our own programs. Those programs were obviously still protected by copyright, even when we typed them up ourselves.
People are playing propaganda games in order to pretend that the copyrights of LLM training material aren't being violated. They're rarely making logical or legal arguments. They just say something about a human learning and a computer "learning," and say since humans learn and own things that they produce based on that learning than computers should be able to "learn," and own things that they produce based on that "learning." It's a motivated word game, not a serious argument.
> A student goes and trains on preexisting work.
point is already shaky. Some of them kind of do?
You have not provided a logical or legal argument for the opposite claim either, and the opposite claim is the one that has the burden of proof in court. In order for LLM training to violate copyright, it needs to fulfill criteria set out by the law that define what constitutes copyright violation. Current copyright law does not mention LLMs directly (unless it's been updated very recently, which I doubt), so a degree of legal interpretation is necessary to determine whether LLM training falls under the law.
It is not wrong (nor illogical!) to point out that such an interpretation must be done with care, lest it suddenly imply that human training is also a violation. We can't have that.
Actual copyright law will generally say that in order to violate copyright, a work must “substantially resemble” the original in some meaningful way (and also not fall under fair use or some other exception), or a similar phrasing that is the same in spirit. Does an LLM resemble any substantial portions of the works it was trained on?
Well, the LLM itself doesn't (it's just a list of weights), but it could be made to reproduce some substantial portions, so you could make an argument similar to your zip file. Such an argument would only apply to situations where you're disseminating the weights, i.e., open weights models.
The case of a chat interface that lets you use an LLM with “secret” weights is a whole other issue. Now the LLM itself is no longer relevant, but only its output (as that is what you're disseminating). Here, the big vendors have taken substantial steps to prevent the output of substantial portions of copyrighted material, even erring on the side of false positives (better to refuse than to risk infringement). Now it could easily be argued that the training process is completely irrelevant because it happened entirely in private, and only the output is relevant and must be decided on a case-by-case basis.
These are some legal arguments, and as you can see, they do not easily or quickly lead to a clear-cut resolution.
I suspect this applies more to audio/visual than it does text or code.
Model vendors know this isn't fair use, given that they're self-censoring the training and/or generation of copyrighted materials by entities big enough to hurt them. As usual, the rest of us have no recourse other than bend over and take it.
> Model vendors know this isn't fair use
Pick a lane dude. Either it's possible for us to make legal analysis and best guesses or uhh "we have no idea it's up for the courts to decide!"
Yes we know it's up for the courts to decide. And people are making argument about what those courts will do.
That's how legal analysis works.
Yes, it is perfectly possible to talk about fair use and follow existing guidelines on it and the government has already make statements that this stuff is likely transformative.
Incorrect.
"…the [Copyright] Office rejects two common arguments about the transformative nature of AI training. As noted above, some argue that the use of copyrighted works to train AI models is inherently transformative because it is not for expressive purposes. We view this argument as mistaken. […] Nor do we agree that AI training is inherently transformative because it is like human learning. To begin with, the analogy rests on a faulty premise, as fair use does not excuse all human acts done for the purpose of learning." — Copyright and Artificial Intelligence, Part 3: Generative AI Training, United States Copyright Office
And again, the degree to which training and output is transformative is just one of several factors that courts consider when deciding if something is protected as fair use.
Several points in that opinion/report (not a policy document) from the USCO also seem to be contrary to rulings like https://en.wikipedia.org/wiki/Authors_Guild,_Inc._v._Google,....
Correct, it can't be as I said above. That why "this stuff is likely transformative" is meaningless in at least two ways.
Fortunately I made no reference to these 2 very specific arguments. So my point stands.
A court has deemed this fair use. The transformation is obviously mechanical, though, unless you think computers are people.
That means I can't know if my AI generated code can be copyrighted unless someone takes it to court?
• What about non-ai code generators (such as https://heyapi.dev/)?
• And what about a code generator that is a hybrid? (So: Something like heyapi but with LLM/agents intertwined)
By which I mean deterministic outputs, not necessarily deterministic processing.
Like if a competitor literally copied all of Spotify’s AI generated music and offered it for free, Spotify would have no recourse.
Maybe Spotify doesn’t care, but it’s something I hadn’t considered.
Nobody would Napster and steal that music. And people can generate similar music for very cheap. Nobody will license that music, say, to use on their game or movie soundtrack.
Spotify only uses it to water down costs for the many eyeballs, or rather earlobes, they have.
This is the part that needs to be defined either via court ruling or legislation. What constitutes "purely AI-generated material"? If I ask claude code to write a new OS kernel, and then write one comment on one utility function, or change a single variable name, is the code mine? Patent Law lacks a fine-grained model for defining ownership of code bases which are often a plethora of licenses and ownership.
We could go one step further and say, instead of changing a variable name or a comment, we alter by hand a for loop to use a recursive function call instead. Do I own the entire OS Kernel now, or does Patent Law have a framework that specifies I own that particular function, but not the entire kernel?
>Copyright does not extend to purely AI-generated material, or material where there is ***insufficient*** human control over the expressive elements.
Vagueness alert.How do we know if it's "sufficient?" Sufficient for what? If the answer is "sufficient for copyright protection," isn't that circular reasoning?
If the answer is "sufficient for an economically successful creative output," isn't this automatically a losing battle? As AI gets better and as movie studios use AI more, the goalpost of "sufficient" will move like sand shifting beneath our feet.
You're confusing writing code with software engineering. Devs never just wrote code, they took ownership of the outcome. Business owners put their money on the line for product decisions.
To say that just because devs or founders use AI to write code, the code is no longer their intellectual property is ridiculous. Humans take responsibility for the outcome and are willing to accept the losses. AI can't do that. AI doesn't have any stakes in the output it generates, humans do.
I know that many people would disagree with me on this, but thankfully I'm free to choose what license I consider "free".
Then you shouldn't release it under the MIT license, which has restrictions and requirements on speech. You have to release it to the public domain. Pretending MIT doesn't dictate terms under which you can release your own software is just a fantasy.
Showing 400 of 665. Read the rest on Hacker News
Comments are loaded live from Hacker News and are not stored by Mid or Real.