Training AI models on copyrighted works is "squarely what transformative use is," EFF’s Tori Noble told Reason - these are "general purpose tools" and "exactly the kind of creation that copyright is designed to foster.” https://reason.com/2026/09/04/doj-says-barring-ai-training-on-copyrighted-material-could-severely-hamper-tech-progress/
180
Replies
0
Boosts
@eff Scale matters. Intent matters. A world-eating model, that benefits from every sentence ever written, cannot morally be locked up behind a paywall for the sole benefit of the few.
@eff interested to see what @pluralistic has to say about this…
@eff A technology only available to ultra-wealthy corporations that is built through copyright infringement is A-OK, but normal people can't have a a 4 second clip of Disney in the background of a video without getting sued.
Is that really what you're saying? I just want to be clear before I cancel my 25-year membership.
@darkghosthunter @Crell @eff No, it is not. LLMs are computer programs that are fed very large sets of data, which they then break down into smaller parts, copy those parts, and extrude the copied parts as plausible, but intentless, sets of text. They simply do what they have been programmed to do, by the people who create them and choose their training data. They do not possess abstract characteristics like "understanding" or "realizing" or "meaning".
@lauerhahn @darkghosthunter @Crell @eff
> They do not possess abstract characteristics like "understanding" or "realizing" or "meaning".
Two thoughts on that
1. I am not so sure, any more, you are correct. These have text prediction at their core, but they seem to simulate two of those three. Turing.
2. I am not sure it is relevant. Do IP rights require understanding?
We are at the beginning of something huge.
Poor us
Not even close to the same. Unless you believe human beings don’t have actual lives and lived experiences, from which the vast majority of their understanding of the world derives.
LLMs are not alive, and have no concept or experience of life. They are probabilistic copying machines. People are more than that.
Granted, creativity is still hard. But people can be creative, with enough dedication. Machines cannot be creative, in any meaningful sense. Art is more than samples and remixes. Even though those are essential parts of it. Artists and audiences have a complex and nuanced understanding of those techniques, and of the context in which they happen.
People know what an allusion is; they know what satire is; whereas an LLM does not. Machines don’t know anything. LLMs are machines designed to plagiarize, and disguise it, while pretending to do something else.
@8r3n7 @darkghosthunter @Crell @eff
> Art is more than samples and remixes
This artist challenges your right to define what art is not
Carry on
@eff I'm so confused. In front of me is Cindy Cohn's "Privacy's Defender". The opening pages read:
"(c) 2026 Massachusetts Institute of Technology
All rights reserved. No part of this book may be used to train artificial intelligence systems or reproduced in any form [...]"
What kind of a f*cking pro-slop-technofascism stance did you assume over night? I'll be dropping my recurring titanium donation if you do not clean this shit up.
@eff have you lost your fucking minds
@eff you what???? :///
@eff well this is the heel turn of the decade, so disappointed to see you siding with the fash
@eff why the fuck are you siding with fascists?
@eff Yeah, nah.
"It's a bold strategy, Cotton. Let's see if it pays off for him."
Yes and reading a book is transformative to my knowledge and mind, so just give me all the copyrighted works. Does anyone dispute that a human reading those works makes a human better/smarter/more creative?
I'm not a huge fan of copyright holder lawsuits. However, AI companies are stacking cash, while standing on the heads of all of human history. It's not a public works project for the betterment of society, it's not a library you can freely use. It's not even like the beginning of the internet, where infrastructure and networking allowed collaboration and content creation.
Love the EFF but I don't think I'd help the AI overlords in this. Spend the time on something better than this.
Heaven forbid that the courts set a precedent that OpenAI pay for content they use for training purposes. Maybe they could just let them use torrent like Meta did already. I mean it's accepted use right. (/s)
@eff Tear out your lying, duplicitous tongues.
You betray everyone who ever looked to you, you goddamn demons.
@eff What a betrayal of content creators. The EFF might as well remove all its humans and function as an AI bot.
