AI Policy for Developers and Contributors
This is our AI Policy for human contributors about using AI assistance with details on:
Legal/IP requirements
Disclosure obligations
Authorship warranties
For instructions to AI assistants (e.g.
Anthropic Claude,
Google Gemini,
OpenAI ChatGPT,
Microsoft Copilot or other)
about providing AI assistance, please see
CLAUDE.md.
Important
The Bottom Line - Make the right thing the easy thing.
Your pull request is your byline. When you submit code, you are signing your name to it and warranting that you are the Author.
Whether you used an AI as an Editor or a Ghostwriter, the final work must be a product of Your skill, judgment, and creative labor.
An AI can never be credited as an author or co-author.
A Helpful Analogy: The AI as an Editor or Ghostwriter
To understand the role you, the human contributor, must play when using AI tools, it’s helpful to draw an analogy from the world of writing and publishing. For any code you submit to be legally sound and protect the integrity of this project, your use of AI must mirror the relationship between an author and their editor or ghostwriter.
Scenario 1: The AI as an Editor
In publishing, an author writes the original manuscript—the story, the characters, the plot. They then give it to an editor. The editor doesn’t write the story, but they refine it. They correct grammar, improve sentence structure, point out inconsistencies, and suggest better ways to phrase things. The final work is stronger, but the Author is, without question, the person who wrote the original manuscript.
This is a perfect model for acceptable AI use:
You write the original, substantive code: the architecture, the core logic, the solution to the problem.
The AI acts as your editor: it helps you refactor for clarity, finds potential bugs (typos), suggests more efficient syntax, or helps write comments and documentation for your code.
In this role, the AI is a powerful tool that polishes your original work. You remain the sole, undisputed Author.
Scenario 2: The AI as a Ghostwriter
A ghostwriter is hired to write a book for another person, based on detailed outlines and extensive interviews. The public figure who hired them takes the byline and is considered the author. They are also fully responsible for the book’s content, accuracy, and any legal issues that may arise. They don’t just blindly publish the ghostwriter’s draft; they must review, revise, and ultimately approve every word, making it their own.
This is also an acceptable, but more demanding, model for AI use:
You act as the commissioner: you provide a detailed, specific prompt to the AI, outlining exactly what you need a block of code to do.
The AI generates a “first draft” of that code block.
Your critical responsibility is to act as the final author: you must meticulously review, test, debug, and often significantly modify that code. You must understand it completely and integrate it into the larger project. By doing so, you take full accountability for it.
In this role, you are using the AI to generate raw material, but it is your intellectual labor — your review, validation, and integration — that transforms it into a valid contribution.