> For the complete documentation index, see [llms.txt](https://www.aicontrolarchitecture.org/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://www.aicontrolarchitecture.org/project/contributing.md). # Contributing The AI Control Architecture is open and openly licensed. It aims, over time, to become a shared standard for AI control; for now it is an openly developed body of work anyone can adopt. It improves fastest when the people using it, security architects, risk and audit teams, privacy teams, and builders, feed back what works and what is missing. Contributions are welcome. *** ## What belongs here This repository holds the AI Control Architecture and its supporting assets: ``` docs/ the architecture: foundation, the ten pillars, operating model, glossary mappings/ crosswalks to external standards and regulations templates/ reusable intake, tiering, control, assurance, evidence, incident templates examples/ worked use-case examples ``` Good contributions include: corrections and clarifications; new or improved standards crosswalks; additional templates and worked examples; sharper requirements; and better explanations of existing controls. ## What does not belong here The AI Control Architecture is deliberately separate from any one implementation of it. Please do not contribute: * product-specific, vendor-specific, or tool-specific instructions; * assessment scoring logic, risk-tiering thresholds, control-selection rules, or pricing, these are operational choices each adopter or vendor makes; * anything that presumes a particular commercial offering. The AI Control Architecture says *what* good AI control looks like and *how to evidence it*. It stays neutral on *whose product* you use to get there. *** ## How to propose a change * For small fixes (typos, broken links, clarifications), open a pull request directly. * For anything that changes a requirement, a pillar's scope, or the requirement identifiers, open an issue first so the change can be discussed before the work. * Keep pull requests focused on one thing. Explain the *why*, not just the *what*. * Run the link checker before submitting: `python3 scripts/check-links.py`. It must report OK. ### Style The docs follow a deliberately plain house style. Match it: * short, declarative lines; prose over decoration; * fenced ` ```text ` blocks for structured lists and diagrams; * ASCII box-drawing for diagrams, not images, where practical; * no em-dashes (use commas, colons, or separate sentences); * every doc ends with a `**Next:**` pointer and the version/licence footer line. ### Requirements Requirements are normative and identified as `ACA--`. Identifiers are stable: do not renumber or reuse a retired ID. To add a requirement, use the next free number in that pillar. To retire one, mark it rather than deleting the number. *** ## Licensing of contributions By contributing, you agree that your contribution is licensed under [CC BY 4.0](/project/license.md), the same licence as the rest of the architecture, and that you have the right to contribute it. You retain copyright in your contribution; you grant everyone the rights CC BY 4.0 provides. Contributions do not grant any rights in the project name or conformance marks (see [TRADEMARK.md](/project/trademark.md)). *** ## Stewardship and governance The AI Control Architecture is stewarded by Neo Control (neocontrol.ai). Stewardship means keeping it coherent, versioned, and moving: reviewing contributions, arbitrating changes to requirements and structure, and maintaining the crosswalks. It does not mean owning the text, that is open under CC BY 4.0. * **Versioning** follows semantic versioning, tracked in `VERSION.md`. Editorial fixes are patch releases; new requirements or templates are minor releases; changes that alter the meaning of existing requirements or the pillar structure are major releases and are announced. * **Decisions** on requirements and structure rest with the stewards, informed by issues and discussion. The bias is toward neutrality, evidence, and keeping the AI Control Architecture usable across vendors and sectors. * **Neutrality** is a design goal. The AI Control Architecture must remain something a competitor of Neo Control can adopt in good conscience. Questions about stewardship, conformance, or use of the name: neocontrol.ai. *** *Version 0.1.0 · Architecture licensed under* [*CC BY 4.0*](/project/license.md) *· Stewarded by Neo Control · neocontrol.ai* --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://www.aicontrolarchitecture.org/project/contributing.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.