For Developers

Proving Authorship of an AI-Assisted Repo

By Nora Whitfield · August 21, 2026

If you think your Git commit history proves you wrote your code, you're walking into a trap. Copyright law cares about human spark, not the automated logs of a terminal. When you use AI to help you build a repo, you're blurring the line between a tool and a creator. If a competitor tries to clone your project or a buyer questions your ownership during an acquisition, they won't ask for your GitHub activity graph. They'll ask for proof that a human being made the creative choices.

The Human Authorship Requirement

The United States Copyright Office holds a hard line on this. In the case of Thaler v. Perlmutter, the court affirmed that copyright law protects only works created by human beings. Machines don't have rights. When you file for a copyright registration, you pay a filing fee of $65 for a standard application. You're swearing under penalty of perjury that you're the author. If the Office finds out you used AI to generate the bulk of your code, they'll reject your application.

This isn't theoretical. Look at the Zarya of the Dawn decision. The Office granted copyright for the text and the layout of the comic book, but they stripped the protection from the AI-generated images. They determined that the AI didn't follow the author's instructions in a way that amounted to creative control. For developers, this means you need to treat your repo as a documented history of human decisions. You aren't just writing code. You're documenting the process of how you guided the machine.

Beyond the Commit Log

Git is great for version control, but it's terrible for proving intent. A commit message that says "update logic" tells a court nothing about how much of that code was prompted and how much was authored by your own hands. You need a paper trail that shows your creative thumbprint.

Start by keeping your prompt logs. When you use an LLM to write a function, save the specific prompts you used. Keep the iterations. If you ask a model to write a sorting algorithm and it gives you garbage, then you rewrite the logic to make it work, that rewrite is your evidence. You're showing the transition from machine output to human-authored code. This evidence chain proves that the machine was a tool, like a compiler or a linter, rather than the primary author.

You should also maintain a design document that details your architectural choices. If you decided to use a specific data structure because it fit your project requirements, write that down. If you manually refactored a block of code generated by a bot, record the date and the reasoning. This isn't just about the code. It's about the thinking that surrounds the code. The Copyright Office looks for evidence that you exercised creative control. Showing that you had a clear plan and that you actively steered the AI toward a specific result is the best way to satisfy that requirement.

The Cost of Silence

If you ignore this, you're leaving your work vulnerable. Statutory damages for copyright infringement can range from $750 to $30,000 per work, and they can climb as high as $150,000 if the court finds the infringement was willful. If you can't prove you're the author, you can't sue for those damages. You can't stop a competitor from stripping your name off the repo and selling your work as their own.

You need to act within the 3-month window after publication to preserve your right to claim statutory damages. If you wait until you're sued or until you're selling your company to figure out your proof, it's already too late. You can't reconstruct your intent after the fact. The time to document your creative process is while you're actually building the repo.

Building Your Evidence Chain

Start a separate documentation folder in your repo. Call it evidence or provenance. Inside, put the raw transcripts of your AI sessions. If you're using a tool that integrates AI directly into your IDE, look for ways to export your chat history or your interaction logs. Do not just commit your code and walk away. Commit your process.

If you're worried about trade secrets, anonymize the prompt logs. You don't need to show the world your secret sauce. You just need to show a third party or a judge that a human made the final calls on the architecture, the implementation, and the refinement. This is the difference between a collection of machine-generated text and a piece of intellectual property you own.

This is not legal advice. Copyright rules change. You should always confirm the current standards at copyright.gov before you file. The Office is still figuring out how to handle AI-assisted works, so keep your documentation flexible and thorough.

Your code is your livelihood. Treat your proof of authorship with the same rigor you apply to your unit tests. If you don't build a chain of evidence, you're essentially building a house on land you don't own.

To start building your evidence chain, download the Copyrightable audit tool to automatically log your AI interactions in a format suitable for registration. Read our methodology guide to understand how we map human intervention to copyright eligibility.