For twenty years, the entertainment industry told you that downloading a movie was theft. They sued teenagers. They sued grandmothers. They ran those anti-piracy ads before the DVDs you already paid for. They lobbied governments to shut down websites, extradite operators, and make examples of individuals who shared files with strangers on the internet.
The message was clear: copying someone's work without permission is wrong. It doesn't matter that nothing physical was taken. It doesn't matter that the original still exists. The act of reproduction without a licence is the crime. That was the line.
And then AI companies copied the entire internet, and suddenly nobody important cared anymore.
Anna's Archive. Library Genesis. Books3. The Pile. Common Crawl. Datasets built from millions of copyrighted works, scraped without permission, without payment, and without so much as a notification to the people who wrote them. Entire libraries of human knowledge and creativity, ingested by models that now generate text, code, images, and music on demand.
When a kid in 2005 torrented a Metallica album, the recording industry called it an existential threat. When a company worth billions trained a model on every book ever digitised, it was called innovation. The difference isn't legal. It's financial. The kid didn't have lobbyists.
Let's be honest about what happened. The same legal system that fined individuals thousands of dollars for sharing a handful of MP3s has, so far, largely failed to hold AI companies accountable for scraping on a scale that makes Napster look like a school library. There are lawsuits in progress, sure. But the models are already trained. The data is already consumed. The products are already shipping.
This isn't a grey area. It's the same act at different scales. An individual copies a file and shares it: crime. A corporation copies billions of files and monetises the output: disruption.
You can't have it both ways. Either copying without permission is wrong, in which case the entire foundation of modern AI is built on something wrong. Or it's acceptable when the result creates enough value, in which case every torrent site operator who ever got prosecuted deserves an apology.
Here's where it gets genuinely strange. AI can now generate passable code, serviceable prose, functional designs, and decent music. Not great, usually. Not human, often. But good enough for a lot of use cases, and getting better fast. The marginal cost of producing content has collapsed towards zero.
That was supposed to be the promise of the internet all along. Information wants to be free. Creation tools in everyone's hands. The democratisation of knowledge and culture. We just didn't expect it to happen by feeding every creator's work into a machine and then competing with them using their own output.
So where does that leave us? I don't fully know. And I'm suspicious of anyone who claims they do.
If you're a writer, your training data is being used to build a system that makes writing cheaper. If you're a programmer, same. If you're a musician, same. The tools are incredible and the implications are terrible and both of those things are true at the same time. That's the part nobody wants to sit with.
I think the torrent kids were right, philosophically. Not because piracy is victimless, but because the principle that was used to crush them is the same principle that's now being ignored at industrial scale. If sharing culture is wrong, then training on it without consent is worse. If training on it is fine, then sharing was always fine too. Pick one.
I think the value of human-made work isn't going away, but the economics of it are broken in ways we haven't figured out yet. The answer isn't to pretend AI doesn't exist. It's also not to hand-wave the theft that built it.
I think we need better answers than "move fast and break things" from the companies and better answers than "ban it all" from the critics. I don't have those answers. I just know the current situation is a mess, and the people who built that mess are the last ones who should be trusted to clean it up.
In the meantime, if anyone from the RIAA wants to explain why downloading an album was a federal crime but ingesting every album ever recorded is a business model, I'll leave the comments section open.
There is no comments section. This is a static HTML page. You'll have to yell into the void like the rest of us.