AI Assistance Policy
How AI tools take part in building our software, what that excludes, and what we ask of contributors.
1. Overview
Polycentric Labs builds its software with AI tools as assistants. This page says what that means in practice, what it excludes, and what we ask of contributors. It applies to every project in our GitHub organization and is mirrored there as AI_POLICY.md.
2. Principles
A person is accountable for every change. The maintainer who lands a change understands it, can explain it, and answers for it. AI output is input to that work, never a substitute for it.
Review before publish. Nothing AI-drafted ships unread: code, tests, documentation, changelog entries and release notes are all reviewed and edited by a person first.
Humans hold the publish keys. Tags, pushes to protected branches, merges, package and container releases, and public statements are performed by a person, on purpose, one action at a time.
3. Disclosure
Projects record their own practice in a dated docs/ai-assistance.md: the disclosure level, the tools in use and their roles, and anything the project rules out. Evidentia's is the first; the other projects follow. Those files are the project-level record; this page is the shared policy.
4. Contributors
Contributors may use AI tools. The contributor is the author and is accountable for the change. Disclose significant AI assistance in the pull request description or with an Assisted-by: <tool> commit trailer.
5. Questions
Open an issue on the project concerned, or write to contact@polycentriclabs.com.