Policies & Guidelines¶

These apply everywhere Aurora Silicon operates — our Discord, our repositories, and anywhere else the project's work is discussed. They are deliberately short, and they are subject to change.

Nearly all of it reduces to one idea: do not bring material into this project that you do not have the right to bring, and do not carry our work somewhere it should not go. For the mechanics of actually submitting a change, see Contributing.

Follow the law¶

Do not break the law. Most of the rules below are essentially this, particularly where commercial, proprietary, or confidential material is concerned.

Apple and Microsoft employees and contractors¶

If you work for Apple or Microsoft — including as a contractor or subcontractor for either company — do not contribute code to this project or to any of our repositories unless both of the following are true:

• You have been explicitly authorised to do so by your employer, and
• An Aurora Silicon administrator has reviewed satisfactory proof of that authorisation.

Confidential and non-public material¶

Do not share NDA-protected or otherwise non-public software, code, documentation, reverse-engineering work, or other confidential material. Public betas and publicly available developer betas are excluded from this restriction.

Any contribution resulting from unreleased, confidential, or otherwise non-public code, software, documentation, hardware, or products will result in the immediate removal of the content, and may result in the accounts involved being banned. Depending on the nature and severity of the material provided, we may also disclose the incident to the appropriate parties.

LLMs and generative AI¶

If you use an LLM — ChatGPT, Claude, GLM or any other — you remain solely responsible for everything you produce and submit with it. The tool does not carry the responsibility; you do.

Make sure your agent follows the AGENTS.md instructions carried throughout our repositories. Those instructions are what keep our requirements around code provenance, licensing, and repository cleanliness intact, and an agent that ignores them will produce work we cannot accept.

Every rule on this page applies unchanged when an LLM is involved. Using open-weight, uncensored, "abliterated", or otherwise modified models exempts you from none of it.

Disclosing AI-generated contributions¶

Any submission containing content partially or fully generated by an LLM or other generative AI must clearly disclose this in the associated commit. The disclosure must be clear enough that maintainers and future contributors can tell AI was involved in producing the submitted content.

This applies regardless of how substantially the generated output was subsequently reviewed, edited, rewritten, or incorporated. Failure to disclose AI involvement may result in the contribution being rejected or removed and, where appropriate, further action being taken by the project administrators.

If an AI generated it under your direction, treat it as your contribution and your responsibility — and disclose it.

Do not contact upstream projects¶

Do not, under any circumstances, report bugs, submit issues, open pull requests, make suggestions, or otherwise send project-derived findings to upstream repositories or their maintainers.

Our approach to LLM usage and decompilation differs significantly from both Asahi Linux and NT-for-ASi. Content produced by Aurora Silicon — or produced as a result of work conducted within this project — should be treated as unusable from the perspective of those upstream projects.

This is not about credit or secrecy. We depend heavily on the projects listed under Attribution, and this rule exists to protect their work from contamination by ours — not to keep ours to ourselves.

If you are unsure¶

Ask an administrator before you act, not after. Every rule here is easier to follow in advance than to unwind later: content that should not have been shared cannot be un-shared, and a contribution built on material we cannot accept has to be removed no matter how good it is.