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/
18
Replies
14
Boosts
Oof. Not the takeiI would have expected from yoy.
The fact that a well crafted prompt can pull near identical data from the training source shows this is not the case.
Touch grass, @eff, I think you have spent too much time online.
@eff Ah. Okay. *writes down EFF on my company shitlist*
@eff I just read the article
I agree
@eff
Wow, the libertarians at EFF bootlicking the techno-fachists.
Not surprised at all.
@eff nope it's theft
@eff Ok, so I’ll download copyright material to train my own actual intelligence? That’s fine right and fair use? Nobody will sue me for breach of copyright, right? RIGHT?
My goddess give your head a wobble. Supporting mass theft of people’s work to feed to the Great Fascist Hallucination Machine? That’s probably the most stupid thing I’ve heard. To hear it from the EFF? Beyond stupid.
This comment is unavailable — it may have been deleted or is not visible to you.
@jcurioso @eff I am curious. How do you think this destroys the commons?
I assume you are referencing the enclosures (https://en.wikipedia.org/wiki/Enclosure)
This seems different to me. Copying does not destroy.
Can you elucidate please?
@eff How about those other three factors of fair use?
@eff thanks for spitting in artists faces....
@eff how to loose everything in one single statement.
@eff lawyers gonna tell you how they see it. Doesn't make it right or wrong.
I do not think fair use applies here. This also doesn't touch native rights at all. One corner of law in very narrow interpretation might be correct, it wider context it is still wrong.
This is usually where new laws get written tho.
@eff
> Training AI models on copyrighted works is "squarely what transformative use is," EFF’s Tori Noble told Reason
People are very upset about this. But it's not a simple 'pro-AI' vs. 'anti-AI' question. It's a complex legal one about where the boundary lines between fair use and copyright violation ought to lie. One can be vehemently opposed to #MOLE Trainers (as I am), without necessarily wanting the copyright monopoly strengthened in the way Big Content is pushing for here.
(1/?)
It's analogous to the link tax issue. Would I feel sorry for DataFarmers if they had to pay individual journalists a royalty whenever their work was linked on DataFarming platforms like FarceBook or Xitter? Not for a second.
So do I support link taxes as policy? Also, not for a second. Not because of their affect on the DataFarmers targeted, but because of the precedent created. If linking required a copyright license, and perhaps a royalty, that would fundamentally break the open web.
(2/?)
Let's say courts ruled that watching/ hearing/ reading a lot of official Star Trek works, then writing your own original fiction set in the Star Trek universe, was a copyright violation. I think most of us would agree that banning fan fiction would be an overextension of the copyright monopoly, so we would oppose this.
The EFF lawyer's take appears to be based on the belief that #MOLE Training is an automated version of this kind of process. If she's correct, her take is reasonable.
(3/?)
However I've seen evidence that suggests generative models are not transformative use. That what they actually do - at least when it comes to source code - is spit up complete copies of data they've ingested during #MOLE Training. Stripped of attribution, and claimed as a new work.
Which is plagiarism, and quite clearly a copyright violation. Not something the EFF ought to be defending. So the correct legal take here is highly dependent on what the tech is actually *doing*.
(4/4)
@eff You just died for me.
