OpenAI declares AI race “over” if training on copyrighted works isn’t fair use | National security hinges on unfettered access to AI training data, OpenAI says.
https://arstechnica.com/tech-policy/2025/03/openai-urges-trump-either-settle-ai-copyright-debate-or-lose-ai-race-to-china/
24 Comments
From the article: OpenAI is hoping that Donald Trump’s AI Action Plan, due out this July, will settle copyright debates by declaring AI training fair use—paving the way for AI companies’ unfettered access to training data that OpenAI claims is critical to defeat China in the AI race.
Currently, courts are mulling whether AI training is fair use, as rights holders say that AI models trained on creative works threaten to replace them in markets and water down humanity’s creative output overall.
OpenAI is just one AI company fighting with rights holders in several dozen lawsuits, arguing that AI transforms copyrighted works it trains on and alleging that AI outputs aren’t substitutes for original works.
So far, one landmark ruling favored rights holders, with a judge declaring AI training is not fair use, as AI outputs clearly threatened to replace Thomson-Reuters’ legal research firm Westlaw in the market, Wired reported. But OpenAI now appears to be looking to Trump to avoid a similar outcome in its lawsuits, including a major suit brought by The New York Times.
“OpenAI’s models are trained to not replicate works for consumption by the public. Instead, they learn from the works and extract patterns, linguistic structures, and contextual insights,” OpenAI claimed. “This means our AI model training aligns with the core objectives of copyright and the fair use doctrine, using existing works to create something wholly new and different without eroding the commercial value of those existing works.”
Oh well, we learned and laughed and made some friends along the way. Okay, not really but I’m happy if someone pulls the plug
In terms of fair use I think he probably has a good argument. Whenever they do mimic something, it’s a sign of poor training, so that might get them over the line
none of this ever crossed my mind about why chinas a.i has been taking off. interesting. good article thank you for sharing
>OpenAI declares AI race “over” if training on copyrighted works isn’t fair use
They’re threatening us with a good time, huh?
In all seriousness though, this is obviously the company being dramatic to save itself from minor-to-moderate inconvenience. Even it became completely illegal to train AI on copyrighted works, and even if OpenAI decided to actually follow that law (good luck making them) then guess what? Everyone would still have a chatbot already trained on a good chunk of human literature and capable of generating copyright-free text almost instantly. People will keep training their bots one way or another, they just might maybe get in trouble for using the laziest and cheapest “let’s take literally all data instantly and without asking” approach possible.
The way I look at AI training is like…
Imagine you go to a lake and fish. Should that be allowed? Maybe – but that’s a different question than if a giant corporation can trawl the entire lake.
It is possible for something to become unethical and something which should be opposed if it is done at an unprecedented scale using technology not seen before. Regardless of a stance on fair use, generative AI bears unique questions on ethics not super applicable to cases outside of generative AI. You have to ask the separate question – is it ethical to trawl content on an industrial scale to source AI modelling? This feels violating on a level very different than a guy making a meme in his basement
Fair Use is our potential methodology to regulate generative AI at the moment. But we should consider if unique protections should be put in place against AI training, as well, even if it is deemed legal by current fair use laws.
Oh don’t worry Sam the military is training their ai on copyrighted use. It’s just you that they don’t allow.
I don’t know this ramifications of this, but if AI has unfettered access to AI training data, all works that derive from OpenAI (or any AI) should be public domain.
If AI touched its creation, or a determined threshold of assistance in a works creation, it should be public domain.
My race to become a billionaire is over if I can’t break into people’s home and steal their valuables.
Ah. My business needs to steal. Make it legal please. (Teenager being sued by Disney for downloading an MP3 is totally totally different right? /s)
Why is it okay for “AI” corporations to violate copyright at an immense scale, but yet individuals are subject to huge civil penalties for doing so?
I am very willing to compromise.
You can train works on copyrighted material for free, if you’re a non-profit with open source AI.
If you’re a for-profit business who trains AI and/or charges for its use, you should be paying.
And as far as AI works go, things generated purely with AI should not be subject to any copyright restrictions at all and should not able to be legally owned including by big corporations. If Warner Brothers makes a film entirely with AI that film should be entirely unable to be copyrighted and fine to be pirated.
Things that are AI-assisted can be copyrighted, provided they meet a reasonable minimum standard. Which is that there is a “human creator” who is an individual and not a corporation and that the amount of AI-assistance is within reasonable limits. In other words, if you generate an entire movie script completely with AI but change 1 word in it, it counts as entirely AI generated. If you write a movie script yourself and have the AI give feedback or make adjustments, it counts as AI-assisted.
All of this means we can continue to develop AI in ways that will help, without screwing over regular artists, writers, actors, etc. just so Warner Brothers can make twice as many billions of dollars by not having to pay creatives anymore.
Cool how about we start with making all code and financials from all of the companies these assholes have been involved in public domain so we can all train on that.
“The laws shouldn’t apply to us, the defense of the nation depends on it!”
Ai Salespeople are the absolute fucking worst.
MMW: if AI training on copyrighted works will be declared non-fair use, then this will be used as a precedent to declare humans training on copyrighted works non-fair use.
Boohoo won’t someone think of my poor plagiarism machine that is setting fire to billions of dollars and ruining the environment in the process
Too bad. Kill it if they’re just going to steal from artists.
Extra extra extra to meet length rules
AI trained on copyrighted material should be legally compelled to be open sourced and copyright free. If you’re not prepared to respect the copyright of others, then you should have no copyright in the products you create from violating copyright.
There’s so much investor lizard money sloshing around in this business, just throw a couple billion at the writers. These guys are just worried it might take them an extra year to become richer than Melon
Here is a solution. Companies pay for a license to access copyrighted material.
We obviously need to figure out how the price is set and how the funds are distributed to authors.
Remember when Sam Altman was the cool open source guy who was going to build non-for-profit tech to make the world a better place? Me neither.
Sounds more like openAI wallet security.
National security doesn’t need nor should it depend on AI.
“If you don’t let us steal other people’s property, we’ll never be able to make money!”
– This Guy
Good, we’re all in agreement, then.
Pay up for commercial use to train your *commercial product*, or you don’t get to make the product.
Seems pretty simple to me, this is literally how licensing for anything else works.