Here's a scenario that hasn't happened to you yet. Someone publishes something that sounds a lot like the framework you built three months ago. The same structure. Similar language. Close enough that your stomach drops.
You know you had it first. You remember the session. The late night, the revision that finally clicked, the moment the idea stopped being vague and became yours.
But could you prove it?
Not I remember writing it. Prove it. With a date. With a record. With something an attorney could hold up and say: this existed here, on this date, in this person's words.
Most creators can't. They have the work but not the trail. The finished document but not the documented process that built it. This module gives you the trail.
Let's get the big one out of the way.
Can I copyright work I created with AI?
Yes. The U.S. Copyright Office addressed this directly in their 2025 guidance: using AI to assist in the creative process does not bar copyrightability. The courts backed this up. In Thaler v. Perlmutter, the D.C. Circuit ruled unanimously that AI cannot be an author. Copyright requires a human being behind the creative decisions.
The test is straightforward. One question: who had creative control?
If you made the decisions, shaped the voice, directed the revisions, rejected what didn't work, and built the final piece through your judgment, the work is yours. Legally, commercially, protectably yours. You can register it, sell it, put your name on it, and sue anyone who takes it.
If you typed a prompt and hit accept on whatever came back, those unedited portions aren't protected. Even if you published them. The Zarya of the Dawn ruling made this concrete: the author's written text got copyright protection, but the AI-generated images she hadn't meaningfully directed did not.
The line between protected and unprotected isn't about whether AI was involved. It's about whether a human was doing the creative work.
This is where most people get nervous. Not the legal principle. The application. Was what I did enough? It's easier to see from two ends of the spectrum.
Someone opens ChatGPT. Types "Write a 5,000-word short story about a detective in 1940s Chicago." Hits enter. Copies the output into a Word document. Saves it. Lists it for sale.
Someone opens a session with a character they've been developing for three weeks. They've already written the backstory. They paste in scene notes and ask for a first pass at the dialogue. The AI drafts something. They cut half of it. They rewrite the pacing. They tell the AI "the second line doesn't sound like her, she'd never say that directly." Four rounds later, they have a scene built from their vision, their ear, their judgment. The AI wrote words. They wrote the story.
The honest truth is that most people using AI as a real creative partner aren't even close to the line. They're so far on the human-directed side that the question answers itself. The worry is louder than the risk.
But it's worth knowing where the line is. Not because you're near it. Because confidence comes from knowing exactly where you stand.
Both Claude and ChatGPT let you export your entire conversation history. Full transcripts. Timestamped. Every prompt you wrote, every response you received, every revision you requested. Most people don't know this exists.
Claude: Settings → Account → Export Data. ChatGPT: Settings → Data Controls → Export Data. You'll receive a downloadable archive. Here's what's actually inside it:
Look at what's in there. The idea. The rejection. The redirect. Three minutes of a Tuesday night, and the record already shows a human making decisions, changing direction, shaping the work. Multiply that across every session you've ever run.
Every moment you told the AI no, try it this way or that's not what I meant or go back to the version from twenty minutes ago is documented. It's the most honest record of creative decision-making you've ever had. You just weren't thinking of it that way.
A copyright attorney would look at that export and see a chain of creation. Proof of origin, proof of direction, proof of human authorship - not just the finished product, but twelve revisions and forty decisions, every one dated to the minute. A novelist working in a word processor has the manuscript and maybe some earlier drafts. You have the entire conversation. Request the export. Store it somewhere you control. Your creative archive shouldn't live only in someone else's infrastructure.
Platform exports are powerful, but they come from the same company that hosted the conversation. In a dispute, that matters. The opposing side can argue the platform's records could be modified. That's where third-party timestamps come in: a different class of evidence entirely, and simpler than the name suggests.
You upload your file. The service generates a unique digital fingerprint of it, called a hash - a string that can only come from that exact document. Change one comma and the fingerprint changes completely. That fingerprint gets paired with a certified time source and sealed. The result is a certificate proving your specific file existed at that specific moment. Not your word against theirs. Math.
The file itself never gets stored or read. Your manuscript stays on your machine. What gets recorded is the proof that it was there: a notary stamp that doesn't require a notary to remember you. It can't be backdated. It can't be altered. Ten years from now, anyone can verify it.
Provenance is an act, not a mood. Press and hold the plate to perform one.
No gate here. The module continues either way.
That's the whole mechanic, performed. For real files, two services do it at document scale:
Anchors your file's digital fingerprint into public blockchains at regular intervals. You submit the fingerprint, never the file itself, and the timestamp becomes part of a public record that no single company, court, or government maintains alone. The certificate can be verified independently by anyone, for as long as the blockchains exist.
Your file gets a verifiable digital fingerprint bound to certified clock evidence, and you receive a timestamp certificate you can store alongside your manuscript. The record can't be altered retroactively. DigiStamp's timestamps comply with RFC 3161, the internet standard for trusted timestamping.
The cost for either is minimal - often a few dollars per file, sometimes free for basic use. Five minutes of time. Permanent value. Timestamp before you share: before beta readers, before agents, before you post an excerpt online. The time to create the proof is before you need it, because once a dispute starts, it's too late to build the trail.
And when the manuscript is finished, register the copyright. U.S. Copyright Office, online, forty-five to sixty-five dollars. Registration isn't required to own the copyright, but it's what gives you legal standing to enforce it: the difference between having a lock on your door and having the ability to call the police.
One more thing before you go build your proof stack.
Everything you just learned is real. The exports work. The timestamps hold up. The legal framework protects human-directed creative work. Use all of it. But use it with open eyes.
The same system that trained on other people's words also trained on patterns, rhythms, and ideas that feel like yours. You put things into the soup. Other people put things into the soup. And sometimes what the AI serves back to someone else carries a trace of what you contributed, just like what it served to you carried traces of theirs. That's not a flaw in the system. That's the system.
Not everything that feels uniquely yours started with you. And not every similarity you encounter out in the world is theft. You will, at some point, see something that looks like your work on someone else's platform. Your stomach will drop. You'll want to reach for the timestamps and the attorneys.
Sometimes you should. Sometimes what you're seeing is the soup doing what the soup does. The protection tools are for the work you know you built. The judgment about when to reach for them is yours - and that judgment is worth more than any timestamp.
The law is clear. The exports are waiting. The tools are ready whenever you are. Now the question shifts to something more interesting: what makes the work yours in the first place?