A Ninth Circuit ruling last month just threw another huge log on the copyright law fire, directly tackling how fair use applies when AI spits out content trained on someone else’s work. The decision has execs in Hollywood and Silicon Valley scrambling to figure out their next move, because it blows up old assumptions about who owns what in this new AI-driven world. It’s a big deal for anyone creating or using online media.
Key Takeaways
- The Ninth Circuit just put the focus on how “far-reaching” an AI’s *output* is, not just its training process, when determining fair use.
- Media giants are suing AI developers left and right, arguing that training models on their copyrighted content is massive infringement.
- Courts are now being forced to define what “far-reaching use” actually means for AI, a definition that will determine who gets sued and who’s safe.
- The writing’s on the wall: expect a future of mandatory licensing deals or entirely new laws just to govern AI training data.
- If you’re a creator or an AI developer, you need to get up to speed on these court cases now, or get ready for some expensive legal battles.
Context and Background
The whole idea of fair use has been a core part of U.S. copyright for a long time. It’s the legal safety valve that lets you use copyrighted stuff for things like commentary, news, or research without getting sued. Whether something qualifies depends on a four-factor test, with a big emphasis on how “far-reaching” the new use is and if it hurts the market for the original. But generative AI breaks this whole system.
These models are trained by scraping gigantic datasets from the internet (which are full of copyrighted material) and then they spit out new things that can look and sound an awful lot like the source data. This is exactly why we’re seeing huge lawsuits from media conglomerates. Universal Music Group sued an AI music platform last year for training on its songs, and The New York Times Company is raising hell about its articles being used to train models, a fight that Reuters has been covering closely. It’s a straightforward conflict: AI developers need data to build their tech, and creators want to protect and be paid for their intellectual property.
Implications for Digital Media
Even though the Ninth Circuit’s ruling came from a visual arts case involving AI-generated images, it’s a huge deal for all of digital media. The court made a really important distinction: what matters most is whether the *output* of the AI is “far-reaching,” not just the act of training the model. So, an AI developer can’t just argue that hoovering up data is fair use if the final product their AI generates is basically a cheap copy of the original material. It has to add something new (a new message or meaning). I think a lot of developers completely missed this nuance.
So what does this mean for them in practice? AI devs and the companies using these tools have to get a lot more careful. It’s not enough to check your data sources. You have to look at what your AI is actually *making*. Platforms will absolutely need to start licensing training data or build much better filters to screen the outputs. If they don’t, the legal bills will just keep piling up, which could either kill some interesting projects or force companies to move somewhere with friendlier laws. Public opinion is also a factor. A Pew Research Center report shows people are pretty worried about AI and copyright. It’s all part of the bigger mess around the lack of global rules, which you can read more about in AI Regulation: A Global Policy Vacuum in 2027.
What’s Next
This whole legal area is a complete mess right now, and it’s not getting sorted out anytime soon. Expect more lawsuits, especially as AI gets so good you can’t tell its output from something a human made. Congress might have to get involved. There are already talks happening in Washington D.C. about changing the Digital Millennium Copyright Act (DMCA) to deal with AI specifically, which could mean creating totally new IP categories or mandating how training data gets licensed.
If you’re a content creator, this is a weird time. You’ve got to worry about protecting your work, but there’s also a chance to create a new income stream by licensing your portfolio to AI developers. For the AI companies, the path forward is to stop being secretive and start talking to the people who own the content. Being transparent about where your data comes from is probably the only way to build a sustainable business. The end game will be a messy compromise between the tech industry and IP holders. The knock-on effects are huge, touching everything from child safety online, as this piece on Child Protection: AI’s 2026 Online Safety Revolution discusses.
The bottom line is that these copyright fights are forcing everyone in digital media to pay attention. Both creators and tech companies need to keep a close eye on how these court cases and potential laws shake out, because it’s going to define the rules of the road for years. This uncertainty even affects how we prepare the next generation, a topic covered in AI Education: Why Kids Aren’t Ready for 2026.
What is “fair use” in the context of digital media?
It’s a part of copyright law that lets you use someone else’s copyrighted work without their permission for specific purposes like commentary, criticism, news, teaching, or research. To qualify, your use has to meet certain conditions, a big one being that it’s “far-reaching” and doesn’t just copy the original.
How does AI complicate existing copyright laws?
The main problem is that generative AI models are trained by consuming huge amounts of data, much of it copyrighted. The big legal fight is over whether that training process, or the new content the AI creates from it, is illegal copyright infringement or can be defended as fair use.
What does “far-reaching use” mean for AI-generated content?
It means the AI’s output can’t just be a rehash of the original work it was trained on. To be considered “far-reaching,” the AI-generated content needs to add a new angle, message, or meaning. If it’s just a knock-off that competes with the original, it’s probably not going to be considered fair use.
Are big media companies winning these copyright lawsuits against AI developers?
It’s still early and the results are all over the place. Some early court decisions have gone in favor of the media companies, but many of the biggest cases are still in progress. The courts are basically building the rulebook as they go, so we don’t have clear winners or losers yet.
What steps can AI developers take to avoid copyright infringement claims?
To lower your risk, you should train your models on data that’s in the public domain or get explicit licenses for any copyrighted material you use. You can also try to design your AI to produce work that is clearly “far-reaching.” Being open about your data sources and how your model works is also becoming a must-do to build trust and avoid legal trouble.
