Jack Conte dropped something interesting at SXSW this week that cuts right through the noise around AI training data. The Patreon CEO isn’t playing the anti-tech card. He can’t, really. But he’s asking a question that’s been bugging me for months: if scraping creator content for training is truly fair use, why are AI companies writing eight-figure checks to Disney and Warner Music?
The logic is pretty straightforward. You don’t pay for something you can legally take for free. Either the fair use argument holds water and applies to everyone, or it doesn’t. What we’re seeing instead is a two-tier system where individual creators get their work vacuumed up without consent or compensation, while massive rightsholders get deals with “multimillion-dollar” price tags attached.
I’ve watched this play out across machine learning forums and legal Twitter threads for the past year. The AI companies keep pointing to precedent and transformative use, but then they turn around and negotiate licensing deals. That behavior speaks louder than any legal brief.
The Pattern We’ve Seen Before
Conte frames this as another disruption cycle, comparing it to the iTunes-to-streaming transition or the shift to vertical video. Creators have been through this before. Platforms change, formats evolve, monetization models collapse and rebuild. The difference this time is the scale and speed at which content gets consumed.
Training a large language model doesn’t just reference your work or get inspired by it. It ingests millions of pieces, processes them, and builds statistical relationships that let it produce similar outputs. Whether that counts as transformative use is still being fought in courts, but the technical reality is that these models wouldn’t exist without that training data.
What bothers me is the selective application of fair use doctrine. If you’re an illustrator with 10,000 followers, your portfolio gets scraped. If you’re Getty Images or Shutterstock, you get a licensing agreement. The legal principle supposedly doesn’t change based on who owns the copyright, but the business practice clearly does.
Follow The Money
I’ll be honest, Conte has skin in this game. Patreon wants a cut of those AI licensing deals for their creator community. That’s not inherently bad, it’s actually rational. The platform has hundreds of thousands of creators whose work likely ended up in training datasets. If there’s money flowing from AI companies to rightsholders, Patreon wants to position itself as an aggregator that can negotiate on behalf of its users.
But the underlying argument still holds regardless of Conte’s business interests. The fair use defense looks increasingly hollow when you examine actual corporate behavior. Companies don’t voluntarily pay for things they can legally use for free. Risk mitigation might explain some deals, but not all of them, and not at these dollar amounts.
The AI companies could make this cleaner. Opt-in training datasets exist. Synthetic data is getting better. There are technical paths that don’t require scraping the entire internet and hoping the legal interpretation lands in your favor. But those paths are slower and more expensive, so we get the current situation instead.
What Actually Gets Built
Conte ended his talk with something that resonates: great artists don’t just play back what exists, they push culture forward. That’s a subtle dig at how LLMs work, predicting the next token based on patterns in training data. The models are getting remarkably good at this, but they’re fundamentally interpolating between existing examples.
I think there’s room for both human creativity and AI tools in the future we’re building. The question isn’t whether AI should exist or whether it’s useful, because it clearly is on both counts. The question is whether we’re going to construct that future with some basic fairness baked in, or whether we’re going to let the usual power dynamics play out where individual creators get steamrolled while corporations cut deals.
The tech industry has a habit of moving fast and asking for forgiveness later, then claiming inevitability when people push back. But there’s nothing inevitable about who gets paid and who doesn’t. Those are choices we’re making right now, and watching Patreon’s CEO call it out in a printed manifesto at SXSW suggests the creator community is done accepting the current arrangement as the only possible outcome.