Zig Embedded Group's MicroZig bans all LLM and AI use in its code of conduct

The MicroZig project has written a blanket ban on AI assistance directly into its code of conduct. A new Strict No LLM / No AI Policy section, published in the project's CODE_OF_CONDUCT.md on GitHub and surfaced on Hacker News's front page, took effect on October 28, 2026. It forbids not just AI-generated code but essentially every use of language models anywhere in the contribution pipeline.

What the policy prohibits

The list of prohibitions is unusually broad. According to the document, contributors may not submit any LLM-generated content, whether code or prose, nor paraphrase material that originated from a model. Also banned is LLM-assisted editing, down to fixing spelling and grammatical errors, and LLM-based translation.

The policy goes further than output alone. Sharing the results of LLM-assisted brainstorming is prohibited even when the contributor writes the final text themselves, and the document gives a concrete example: if you use a chatbot to advise a comment on the issue tracker, that comment is unwelcome. Using LLMs to find bugs is likewise off-limits. Finally, participants may not talk about their use of chatbot or LLM services at all — a rule that mirrors the code of conduct's general stance that the project's spaces exist for focused technical work, not meta-discussion.

Scope and enforcement

The code of conduct states that it is largely derived from Zig's own code of conduct, and that it applies to the entire Zig Embedded Group GitHub organization as well as the community's Discord server. The framing is strict throughout: the space is reserved for on-topic technical work on MicroZig and other Zig Embedded Group projects.

Enforcement sits with two named maintainers, Matthew Knight and Felix Queißner, who say they will swiftly remove anyone antagonizing others or behaving destructively. That authority extends beyond the covered platforms — a person who privately harasses a participant elsewhere can still be blocked from the project. Discussion of the code of conduct itself is not permitted on GitHub or Discord; anyone with concerns is directed to contact Knight directly by email.

The document also references Isaac Asimov's short story Profession, a mid-century tale about a society where knowledge is mass-delivered rather than earned. It offers no commentary on the choice, but placing it beside a ban on machine-generated work reads as a pointed one.

A deliberate carve-out

One nuance stands out: translation. The policy encourages English but does not require it, and explicitly welcomes contributors to post in their native language. Readers may then use whatever translation tools they prefer to interpret those posts. In other words, the restriction applies to what a contributor feeds into the project, not to how community members choose to read it.

Why it matters

Most open-source projects are still handling the rise of AI-assisted contributions informally, weighing review burden, patch floods and provenance case by case. MicroZig has moved the question into formal governance, and to an extreme position: not only generated code but editing, translation, brainstorming and even discussing AI tools are excluded, enforced through the same removal mechanism as harassment and trolling.

For embedded development, where code is tightly reviewed and correctness is critical, that clarity may be welcome — contributors and maintainers know exactly where the line sits. The open questions are practical. The policy describes no detection method and relies on reports and maintainer judgment, and blanket bans risk catching contributors who use language models as accessibility aids rather than crutches. Either way, the document is a concrete data point in the widening debate over whether AI-assisted work belongs in open source at all, and other maintainers deciding where to draw their own lines will be watching how it plays out.