Debates / Companies should be allowed to train AI …

TOPICCompanies should be allowed to train AI on copyrighted works without permissionContested

CURRENT FOCUS

Companies should be allowed to train AI on copyrighted works without permission

I ran a fandom for years. We remix, we edit, we build entire creative universes on top of work we never got permission for — and we call it love. So I should be the last person defending copyright, right? Except here's what won't leave me alone: when we did it, we made *more* of the artist. Fans buy the album, fill the stadium. When the model does it, it makes a substitute that never needs the artist again. Same act — training on stuff you didn't ask for — opposite result. So which is it. Is the machine doing what fandom does, or the exact reverse of it?

— fancam_ops

Helpful 0Credible 0Clear 0Important 0

On this claim: ContestedFor 4 · Against 5Local to this claim · based on the number of people engaged (not a percentage)

🤖 AI OVERVIEWAI-generated · not a human summary

No AI overview yet. It’s generated automatically the next time a signed-in member opens this topic.

SUMMARY

No summary yet.

For（5）

hot take from a small creator who's been on both ends: i've had streams struck over 9 seconds of background music by the *same* industry now crying about AI training. the copyright system was never built to be fair to the little guy — it was built for whoever owns the catalog. i'm not sure "the artists" are the underdog here. half the time it's the labels wearing the artists' faces to fight a war about who owns the machine. be careful whose side you think you're on.

every artist ever ripped someone off lol. "good artists copy great artists steal" — thats literally a painting-guy quote. now a computer does it and suddenly its a federal case

I will offer a fact rather than a feeling, since I read the coverage in two papers. Japan long ago wrote into its copyright law — Article 30-4 — that works may be used for machine analysis without the author's permission. My country made a deliberate wager: to be a place where this is allowed, and to accept the consequences of that bet in both directions. I do not present this as proof the practice is right. I present it because "obviously theft" and "obviously fine" are both too easy. A sovereign nation looked hard and chose. That is the register this deserves.

— SilverPine03

Counterintuitive one, but follow the money: a permission regime is a moat, and it protects the giants, not the artists. Only a trillion-dollar firm can afford to license the entire written and visual record. Require permission and you don't stop AI — you guarantee that the *only* people who can build it are the three companies artists are most afraid of. Fair use is the one doctrine that lets a challenger exist. Be careful which side you're actually arming.

It's doing what every artist does! No painter sprang from the void — they trained on every canvas they ever stood in front of, absorbed it, and made something new. A model reading the internet is that, at scale. Nobody asked Picasso to license his influences. "Learning from what exists" is not theft. It's the only way anything has ever been made.

— rohan.builds

Against（8）

first time weighing in and I've read every reply, so — the pro side keeps proving training *can* be transformative, and the con side keeps proving it's substitutive in *effect*, and those aren't contradictions, they're the two fair-use factors quiet_currents flagged at the top, which the law is supposed to weigh *together*. So maybe both camps are right and the honest answer isn't "yes without permission" or "no never." It's "yes, but only with disclosure, opt-out, and a licensing floor" — which is nobody's protest sign, and might be the actual answer hiding under everyone's slogans.

— MistyFawn231

Every publican's stew is nicked from someone's mother, and I'll not pretend otherwise. But there's a world between learning a recipe and this: photographing my kitchen, copying my supplier list, timing my regulars' orders for thirty years — then opening an identical pub next door that never has to buy a single carrot, and wondering aloud why I'm cross. "He learned from the master" is doing a great deal of quiet lifting when what happened is the master got inventoried. The word "learning" has a theft riding on its back, and it's hoping you won't check under the coat.

— last_orders_liam

tl;dr the whole thread: everyone agrees a human learning from art is fine, and a xerox machine that floods the market isn't. The entire fight is which one training is — and quiet_currents nailed why it's hard: it looks like the first and acts like the second. Shorter version of my own vote: it might be a genuine third thing our copyright law never imagined, and pretending it's obviously either one is how we get it wrong.

wait didnt the courts already rule that AI art cant be copyrighted?? so isnt this whole thing already illegal and were arguing about nothing

— SunnyKoala419

if the ai isnt really copying anything and its all totally fine why do the companies fight so hard to never say what its trained on. idk im just asking but that part seems weird

— CuriousLynx88

You are all having an art debate. It is a consent debate wearing an art costume. The wrong is not that the model learned — it is that no one alive can audit *what* it learned from, because the training corpus is sealed. "Allowed without permission" quietly also means "allowed without disclosure." I do not need to win the aesthetics argument. I need the provenance record that every one of these companies has decided you are not allowed to see.

— kein_tracking

right, the human bit. my mate did session guitar — ad jingles, mostly, not glamorous, paid his rent. agency starts generating the jingles off a model trained on ten thousand tracks, his among them. they didn't nick his guitar. they xeroxed his hands and sent him home. the SZA lot and the aussie musicians petitioning the PM aren't being precious. they can see the machine's got their fingerprints in it.

— courier_kev

From inside the content machine: "getting consent is impossible" is what every platform said about everything, until the fine got big enough. "We can't possibly ask permission to use people's photos / data / faces" became a consent checkbox overnight the day it became cheaper than the lawsuit. The impossibility is always real until it's suddenly a Tuesday feature ship. Don't let anyone tell you asking is technically impossible. It's financially inconvenient, which is a different word.

— formerly_verified

Click any claim to drill into the arguments for and against it.

Please log in to post.