In October 2023, the U.S. Copyright Office made something official that had been quietly true for months: fully AI-generated images, text, and art cannot be copyrighted. Not because the law wasn't ready. Because the law decided machines don't create—they produce. That distinction just became your problem.
I talked to Maya Chen, a freelance illustrator and designer in Portland who spent six months testing AI tools for client work before realizing she couldn't actually own what she was making. She tried anyway. Here's what she learned.
So you tested AI as a tool for client work. Walk me through what changed your mind.
Look... at first I thought AI was just Photoshop with more automation. I could prompt, iterate, refine, and deliver the final work to a client. But then a client asked me to license an image for five years, and I had to tell them I couldn't actually guarantee ownership of what I'd made. That's when I started reading the actual Copyright Office rulings.
The thing they're saying is this: if the creative choices—the composition, the direction, the conceptual core—come from a machine learning model rather than human authorship, the Copyright Office won't register it. That means no legal recourse if someone copies it. No grounds to defend it. You can't actually own it, which means you can't actually sell it in any meaningful way. A client needs a license they can trust. I couldn't give them that.
So I went back to drawing. Pencil first, then digital. It took longer. It cost more. The client paid for that. But now they own the work. That matters more than I thought it would.
Did you try the hybrid approach—using AI to generate rough directions, then hand-finishing everything?
I did. And here's the honest part... it depends on how much you actually hand-finished. The Copyright Office looks at what percentage of the work is human authorship versus AI-generated elements. There's no hard line. It's murky. A lawyer told me that if you're spending 70, 80 percent of the time fixing AI output, you're probably in the clear. But "probably" isn't a contract you can sell to a client.
What actually happened is I'd generate something, then spend hours making it mine. And somewhere around hour three, I'd think... why am I paying for AI to create something I have to recreate anyway? The time savings disappeared. I was paying for the tool and doing the work. That math doesn't work for solo operators.
The hybrid thing works if you're using AI to do boring iterations or variations on a locked concept you've already created. Not as your primary authorship tool. Not if ownership matters to your client. And if they're paying for it, ownership always matters.
What about using AI for things that aren't final deliverables—research, moodboards, reference?
That's where it actually works without friction. I use it to explore direction when I'm stuck. I'll generate 20 variations on a concept, grab elements, and build from there. Nobody's paying for the AI output. I'm paying for the tool as a thinking aid, like a mood book or a mood board tool. That feels honest. No copyright questions. No ownership confusion.
The problem is that once you've used it that way, the seduction is real. You start thinking... what if I just keep that one? What if I deliver it as-is? And suddenly you're selling something you don't actually own. Your client is unknowingly buying something they can't defend. That's not a service. That's a liability with a pretty picture.
How did your clients react when you stopped using AI on final work?
Honestly? They didn't care. Nobody's hiring me because of AI speed. They're hiring me because they want work that feels like *something*... like there's a person behind the decisions. The ones who want fast and cheap want it cheap, and I can't compete with that. That's fine. I'm not trying to.
What surprised me was the clients who said... oh good. They didn't trust the AI work. They wanted to know there was a human standing behind the decisions. That someone would be accountable if something went wrong. AI can't be accountable. I can.
Do you think the copyright ruling changes how solo creators should approach their business model?
It already is. The rubric is becoming clearer: you either own your work or you compete on price. Those are different businesses. If you own it, you can license it, defend it, build a catalog that compounds over time. If you don't... you're trading time for money on every single project. That's freelancing, not building.
The copyright trap is that AI lets you scale production really fast, but it doesn't let you scale ownership. You make more things faster, but each thing is worth less because nobody's sure who actually made it. It's a speed trap that feels like growth until you try to sell it.
What would change your mind about this? What would make you use AI for final work again?
If the law changed and full AI generation became copyrightable to the person who prompted it... maybe. But even then, I'm not sure. Because the other problem—the *real* problem—is that everyone gets the same tool. We all have access to the same model. If I make something with AI, so can someone else, in seconds, with a slightly different prompt. What I make isn't rare. It's not defensible. It's not mine in the way that matters.
What makes my work worth owning is that it came from taste, from training, from thousands of hours of decisions that can't be replicated. That's what clients are actually buying. Until AI can't be accessed by everyone equally... I don't see how it solves anything for a solo creator. It just makes the commodification faster.
Final question: are you angry about this? Or just... pragmatic?
I'm not angry. I'm relieved, actually. For about three months I felt like I was supposed to adopt this tool or get left behind. Every newsletter, every designer online was talking about their AI workflow. I felt behind. But then I realized... the people who felt left behind are usually the people who wanted shortcuts anyway. The work didn't feel good without ownership built in. I'm better at the thing I actually do. The tool doesn't change that. It just shows me what I actually care about building.