You Can't Protect What You Don't Know You Own

In Episode 70, I sat down with Patrice Perkins, founder of Creative Genius Law and my business attorney for the last six years, for a fireside chat about who owns work that AI touched, what belongs in your contracts now, how to keep your frameworks yours, and when a bootstrapped business actually needs a lawyer. This post pulls the parts of that conversation I keep turning over, plus a few lessons I paid for the hard way. It's education, not legal advice for your specific situation. Note: this post may have affiliate links and we donate 100% of those funds to non-profits and a children's group home🫶🏾 

Somewhere in most small creative businesses right now there is a folder of work a chatbot has touched. Maybe it's a sales page reworked in Claude at 11pm, or a logo concept someone generated in thirty seconds and liked enough to hand to a designer. Most of us assumed the question of who owns AI-generated content would sort itself out. Or we assumed the answer was obviously us, since we were the ones typing the prompt. The US Copyright Office has been saying otherwise for more than a year, and the courts have mostly agreed. Many of us built our businesses on original thinking: the frameworks, the curricula, the voice clients pay for, the methodology we plan to retire on. For us, the gap between what we assumed and what the law says is where real value leaks out. The honest version of this topic is harder than the productivity-hack version we get fed on LinkedIn, because it asks us to decide what our work is worth protecting before anyone forces the question.

Who Owns AI-Generated Content?

Patrice opened our conversation by calling the current state of AI and copyright the Wild Wild West. She meant that for the legal industry too, which is still working it out in real time. The Copyright Office's January 2025 report on copyrightability drew the line much tighter than many of us expected. It concluded that prompts alone, even long and sophisticated ones, don't make you the author of what comes out. Patrice walked us through the example that made it click for me. Describe a gray cat with spots in particular places and one paw slightly larger than the others. Paste that exact prompt into ChatGPT and then into Gemini, and you'll get two different cats. The law reads that wiggle room as evidence that the machine made the creative choices, and copyright only protects creative choices a human made.

The stakes get clearer once you see what ownership does for you in practice. Owning the copyright is what lets you license your work and control its spinoffs (Patrice's favorite example is the film adapted from a novel). It's also what gives you standing when someone lifts your work wholesale. That protection attaches the moment the work exists in tangible form and in most cases lasts your lifetime plus seventy years. Patrice mentioned reading several recent novels she's fairly sure were machine-written, and the part that stayed with me had nothing to do with the prose. If someone copies one of those books, the person whose name is on the cover can't stop them, because in the eyes of the law they were never the author.

Can You Copyright Something AI Helped You Make?

The answer depends on the role AI played. The line Patrice draws runs between AI as a thought partner and AI as the author of the deliverable. You might ask it to find the blind spot in an email draft, or to play a skeptical buyer poking holes in a service you're about to launch. Either way you stay in the author's chair, since you're still the one making the final thing. Handing it your draft and asking it to make the whole thing better is where her read gets conservative, and she treats what comes back as AI-generated. The Copyright Office report does leave some room here (your own expression that survives into the output can still be protected). That room is exactly the territory courts are still mapping, though, and your memoir is the wrong manuscript to test it on.

The logo question came from the audience, and it shows how this plays out inside a brand. A logo lives under trademark law, which protects customers' ability to recognize your business out in the world. A well-cleared mark carries real weight when a company gets valued for financing or a sale. When someone generates a logo in an AI tool and hands it to a designer as a reference, they're handing over something they never owned. Patrice's advice is to let the designer run their own discovery instead. Even a designer who tries hard to forget your AI mockup will remember pieces of it.

"My chat is Nova. That is Nova's logo. That is not your logo." — Patrice

What an AI Clause in Your Contracts Should Cover

Most of us who sell services sit on both sides of this at once. Clients may be feeding our frameworks into AI tools, and the contractors we hire might be feeding in our brand. Patrice calls this a new contract season. Her fix starts on paper and then gets said out loud. On the service provider side, the contract can state plainly that your frameworks and worksheets may not be uploaded into any AI platform. The onboarding call is where you say it again, because nobody reads the fine print. On the hiring side, she wants contractors to represent that their work is original and to disclose any AI use up front. She'd also add a promise not to upload your mockups or working brand names into third-party platforms. She recently drafted a middle ground for a brand designer who worried that Adobe's built-in AI features would put them in breach of their own agreement. The language allows AI that ships as a standard feature of industry software. It rules out AI-generated brand assets and any uploading of client materials.

She also said something about the client side that I wish someone had told me earlier in this business. Ask a service provider how they use AI. If they get offended, or they shrug and say this is just how things work now without offering any middle ground, that's your warning. One of her clients had spent years developing an original character. She was nervous about asking her animator, worried she'd come across as difficult, and she asked anyway. She got to name exactly where AI was welcome (transitions were fine) and where it wasn't (her character was not going into any platform).

"If they're offended that you asked, red flag. There's always a way to meet you in the middle." — Patrice

I'll add the part that's less comforting. A contract works mostly as a deterrent, which counts for a lot, but it isn't a force field. A breach you never find out about is a breach you can't do much with. I once took a client dispute all the way to litigation and won, and that client still hasn't paid what the court said they owe us. I've made my peace with not spending another small fortune chasing it. What I took from that experience is where my energy belongs: on what goes into the contract and what I never hand over in the first place. Some people will breach no matter how good your language is, but most people read a clear boundary and respect it.

The Work Worth Keeping Out of the Machine

One audience member asked a question plenty of us have quietly wondered about. She'd drafted an event poster with ChatGPT and taped it to her window. Was OpenAI going to come after her for it? Patrice's honest answer was possibly, possibly not, and she said the math changes once real money is involved. She'd be more wary of the original creators whose work trained these tools. In July 2026 a federal court gave final approval to the largest copyright class action settlement in US history. Anthropic agreed to pay $1.5 billion to authors and publishers over pirated books it downloaded from shadow libraries, and some authors opted out to sue on their own. The people with the strongest footing in that fight were the ones who could prove the work was theirs. Patrice's working assumption is to treat whatever you put into these platforms as out of your hands, whatever today's terms of service promise, and she offers that without any paranoia attached. Terms change, and nobody knows yet whether a change would apply retroactively.

Most AI advice for small businesses treats the tools as a race you're losing by the day, and I think that framing is its own kind of scarcity culture. Patrice compared this moment to the early years of Instagram and Twitter. Her advice back then was to make sure ownership could always be traced back to you, and her advice now is nearly identical, with more sophistication layered on top. What she recommends is placing your work on a spectrum of value. A routine letter can go into a chatbot once you've stripped the confidential details. The original thinking you plan to retire on should never go in at all. Everything in between gets a narrow, specific reason for using AI, or it doesn't get AI.

I hold strict boundaries around what I give these tools. Part of that is practical: I built something successful and deeply human before AI existed, and I trust myself to keep doing it. The other part is about the kind of mind I want in my sixties and seventies. I want one that can still sit with a client in a hard moment and think past the first answer instead of settling for whatever a model hands me. I've stopped taking cues from LinkedIn hot takes, which wreck my nervous system and are often simply wrong. Emily M. Bender and Alex Hanna's The AI Con and Karen Hao's Empire of AI have done more for my discernment than a year of scrolling ever could. Both come from people studying the policy and the concentrated power behind these companies. My counseling training drilled into me that fear gets smaller once it gets specific. Knowing exactly what I own, and what I refuse to hand over, is the most specific I've felt about AI in a long time.

How to Protect Your Intellectual Property From AI

Document as you go. Patrice is a big fan of journaling, professionally and personally. Notes on when an idea started forming and when you first shared it or took it to market build a paper trail pointing back to you as the creator. A simple creation log counts, and so do descriptions in your file metadata.

Give your ideas a home you own. A friend of Patrice's shared a new framework with her in person, one likely headed into a pending book. Four hours later it was going viral on Threads, with the friend happily chiming in. Patrice's text opened with "Ma'am." She told her friend to pull the post, publish the thinking as an article on her own website, and come back to Threads with a link. Once a social platform becomes an idea's first home, attribution (your right to be associated with your work) gets easy to lose. That's exactly what a publisher weighs when they get interested in your framework. Share the high-level version from your own site and keep the working parts private, knowing that anything public can still be scraped.

Write an AI usage policy, even as a team of one. It can fit on half a page: what you allow and what you don't. Anything unclear comes back to you for a decision. It works as a checkpoint for you and as an attachment to contractor agreements.

Check the Copyright Office every quarter. Patrice recommends its one-page briefs on recent court decisions, which explain what was decided and why in plain language. If you're outside the US, check whether your country belongs to the Berne Convention (more than 180 countries do, Mexico included). If it does, US developments are a useful signal for where things are heading.

Clear your trademark before you build on it. This is where Patrice says waiting gets expensive. Clearing a name or logo at the start runs roughly $2,300 to $3,000 in her practice. Discovering a conflict years in usually costs at least $10,000 to resolve. Negotiating shared rights to a contested name added about $5,000 for one of her current clients. A trademark opposition typically starts at $15,000 to $20,000. For most of the AI questions above, a quarterly check-in and a limited consult will do. She'd bring in a lawyer when a contractor refuses to address AI in the contract or an ownership question gets genuinely complicated. Her firm, Creative Genius Law, is where I'd start.

There's a thread from this conversation I haven't finished pulling: what an AI usage policy looks like when it's written plainly enough for a new contractor to follow on day one. I'll come back to it in a future post. The AI conversation Patrice described usually starts on a discovery call, well before anyone signs anything, and that's also where most client relationships are won or lost. If you have clients now and want more of those first conversations on your calendar, the scripts in How to Book Your First 10 Discovery Calls are free here

Originally aired as Episode 70 on the Choose Courage Inc. podcast.