The Founders (USA) explicitly included copyright in the U.S. Constitution to strike a balance between allowing creators to benefit exclusively from their works, and having those works in the Public Domain to be exploited for the Common Good.
If anything, the copyright extensions championed by Disney over the decades have done more in the creator’s interests than anything else.
[the United States Congress shall have power] 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.
As I understand it, recent rulings have clarified that rightsholder permission is NOT required to use a publicly available copyrighted work to train an AI. All that is required is that the AI company pays for the book. Rightsholders also cannot “opt-out” from having their book included in AI training, any more than they can “opt-out” from having Republicans read their book.
We do disagree about this. I believe the quality of art is inherently subjective. Some people love Jackson Pollock, others see a drop cloth. Both are right. If AI can make you feel like a creator and think you’ve made something marvelous, that is sufficient. The world need not agree.
I haven’t coded in dog’s years. I never learned javascript. I had a sticking point in a workflow the other day, and I vibecoded the solution in JS with Claude AI. It felt glorious not to be helpless, to close the open loop and solve the problem, and to coach the AI in the parts it couldn’t perceive. Every time I use the JS solution, I get a little smile/dopamine. My world is a little bigger, and I’m emboldened to try slightly more ambitious coding projects.
It’s possible that a senior engineer would take issue with the structure or implementation of the javascript, or that I used JS at all when python would be more elegant or durable. I don’t give a fig. it overcame an obstacle that had me at a standstill, I participated, and now I’m more productive.
Don’t back away. Embrace the bigger future. Claim you superpowers in this wonderful new world.
It’s a “misunderstanding” that, in the US, goes all the way back to the Constitution:
[the United States Congress shall have power] 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.
Obviously a complete discussion of international IP law is beyond the scope of this forum, but courts that have looked at the issue have said that interpreting it to mean “you can ingest enormous corpuses of copyrighted works without compensation to train a model for the benefit of a single corporation” is novel, to say the least.
It didn’t demonstrate that. “Regurgitation” requires storage. You can regurgitate lunch because lunch is still in your stomach. AI training doesn’t store any of the original training material, just measurements and correlations between those measurements. So the demonstrations don’t prove that the AI is storing/copying the copyrighted material, it’s demonstrating that it can be instructed to create an infringing facsimile. The demonstrations even mentioned that when ordered to duplicate a film frame, the resulting frame was not an exact copy, but a close simulation.
You might think it’s close enough, but it isn’t. AI training isn’t infringement, and ordering an AI to engage in infringing behavior is the fault of the human operator, not proof of infringement inherent in AI design.
Jim Cameron has the right idea. He said rightsholders should be concentrating on infringing output, where it hurts their business. Let the machines study everything, but charge the companies for the rights to use your characters. Like the deal WB recently made.
I don’t have a problem with that, when the AI becomes a better coder than the senior programmer. The boffins in Silicon Valley are predicting that will be in 12-24 months. Already, AI can code as well as most senior programmers, and virtually no professional coders are developing without using AI. It is ubiquitous.
It’s more like “the law says you can legally infringe on copyrighted works if your use is transformative; having an AI measure and correlate copyrighted works to create new and unrelated works qualifies for this exception.” There’s nothing novel there, it’s a strong and standard Fair Use ruling.
The AI companies say they don’t store the training material. But people have demonstrated that models will produce very long – chapter length or greater – verbatim passages. That’s infringement.
Yes, of course AI companies argue that creating infringing works violates their terms of service. Just like they argue that they aren’t responsible if users use their model to create CSAM material. We’ll see if they suddenly figure out how to enforce local laws when the alternative is to be banned from various countries altogether.
If you’ve taught IP academically, I’m surprised you’re startled. This isn’t controversial. As I noted above, I’m already over my forum word limit but I’ve put 25c in the meter for another 300 words to quickly respond. I’ll keep it brief rather than inflict a thesis on you; I just want to address two points: motivation and effect.
Motivation:
@popcornflix helpfully quoted the relevant piece of the US Constitution where the very first words explain the rationale. The only reason IP law exists is to: “promote the progress of science and useful arts”. [1] The clause literally explains that IP laws exist to avoid discouraging people from making useful things, with limits to ensure that others can use them.
Effect:
Here’s your homework: Review all the many exclusions, limits and exceptions to IP rights and consider if they feel like they’ve been implemented to protect the creator, or to make sure that protection is no bigger than necessary to keep creators creating?
Then your exam question: What does the response of the world’s Governments to IP issues in AI suggest about their belief about the balance of rights between the individual creator and the maximisation of wider economic benefit?[2]
I’ve heard it argued that the only reason IP is mentioned in the Constitution at all is because IP laws are considered an infringement on your free speech freedom to copy things… although I think that’s quite tenuous myself. ↩︎
hint: Australia’s stance is an excellent data point against which to calibrate other nations ↩︎
@pigfender, I’ve been thinking carefully about your assertion that intellectual property law has never been about protecting creators, but instead about maximizing returns for people who want to copy things. It’s a striking claim, and I understand why you frame it that way, especially given the long history of publishers, printers, and distributors shaping the practical realities of copyright. But when I look at the constitutional foundation of IP in the United States, I see a different starting point. Article I, Section 8 empowers Congress to secure exclusive rights for authors and inventors for a limited time, and it does so explicitly to promote progress in science and the arts. That language presumes that creators have something worth securing and that society benefits when their ability to create is supported.
You’re right that the law has often been influenced by the interests of those who reproduce and distribute creative work. Historically, the people with capital and infrastructure have had disproportionate influence over how IP systems function in practice. But that doesn’t erase the fact that the legal framework was designed to give creators a meaningful period of exclusivity. Without that, the incentive structure collapses. The entire point of granting a temporary monopoly is to encourage the production of new work that eventually enriches the public domain. That is not a creator‑hostile premise; it is a recognition that creative labor has value and that society benefits when that labor is sustainable.
Where I think your framing risks oversimplifying the landscape is in treating creators as incidental to the system. The early history of copyright is full of examples where authors were exploited precisely because they lacked legal protection against the entities that could mass‑produce their work. Copyright emerged in part to correct that imbalance. Even if the practical beneficiaries have often been intermediaries, the legal justification has always been tied to the act of creation itself. The system may not always function as intended, but its purpose is not reducible to maximizing returns for copiers.
The modern environment complicates all of this. Digital reproduction is instantaneous, AI systems can ingest and emulate creative work at scale, and the economic value of originality is under pressure. In that context, the rationale for protecting creators becomes even more important. If the law shifts too far toward unrestricted copying, we risk undermining the very creative ecosystem that makes copying valuable in the first place. If it shifts too far toward absolute control, we stifle innovation and the free flow of ideas. The challenge is not to abandon the original purpose of IP law, but to reinterpret it responsibly for a world where the mechanics of copying have changed dramatically.
So while I understand the argument that IP law has often served the interests of those who copy and distribute creative work — particularly evident in how some governments are balancing creator rights against encroachment by AI companies, as you’ve pointed out — I don’t think that means creators were never meant to be protected. They are the origin point of the entire system. The law exists because society recognized that creative work has value, that this value deserves a period of exclusivity, and that the public interest is best served when creators have both the incentive and the ability to continue producing new work. That is the balance I believe is worth preserving though I am the first to acknowledge that it is hard to get the balance right in this new age of AI.
I probably should point out that I’ve also studied IP law academically to a pretty high level (albeit UK based, although let’s face it there’s nonsuch thing as purely jurisdictional IP issues and context).
I didn’t say that. Creators are absolutely meant to be protected, otherwise they wouldn’t create. But why do we (society) care if they create? So we can benefit from progress in science and useful arts by copying them.
Where the lines are drawn by legislators is based on balancing the interests of both creators and exploiters to create maximum economic benefit to society. When you look at how tightly (for example) copyright protections are drawn, with all their limits, exclusions and exemptions, it’s pretty clear what the guiding principle is.
A good litmus test is to ask if a given hypothesis accurately predicts legislative and judicial response to emerging issues. Government responses to emerging issues don’t look to be on the side of the little guy here.
But really, none of what I’m saying is new or controversial if you’ve studied IP law academically;[1] it’s basically a truism.
·~·
I recently read a (very short) article on how to make the leap to become a professional writer that really opened my eyes. It said: If you want to become a professional, you first have to stop being an amateur.
One amateur thing I need to stop doing is getting sucked into internet debates about things when I should be writing. I am failing, clearly, but am going to try harder.
This is all by way of apologising if I don’t engage with your excellent and insightful responses. My failure to respond doesn’t mean that I disagree with you, agree with you, or don’t care. It just means I’m terribly disappointed with my life and trying to do something about it. Love all of you to bits (yes, all of you).
Stay awesome.
Rog x
with the emphasis on the philosophical, wishy-washy nature of academia as opposed to, you know, actual practical pragmatic jurisdictional legal advice ↩︎
The AI companies are sensibly positioning themselves for Common Carrier protection, like a telephone, internet or social media provider. They rightly argue that they should not be responsible for the output created by users using their tools.
Would you prefer that an angry reader could sue Scrivener over a book you wrote? Tool makers should not be held responsible for the misuse of their products.
Scrivener doesn’t “write” anything. The AI tools do. That’s the entire point.
As for Common Carrier protection, “social media providers” and “telephone providers” are treated very differently under the law, so I’m not sure what you’re saying here.