MicroZig, an open-source toolbox for building embedded applications in the Zig programming language, has adopted a strict policy barring content created or assisted by large language models from its project spaces. The rule appears in the Zig Embedded Group’s code of conduct and applies across the organization’s GitHub repositories and Discord server.
The policy covers more than generated source code. It prohibits LLM-produced prose, editing assistance for spelling or grammar, automated translation and the sharing of ideas developed through chatbot brainstorming. It also says that advice obtained from a chatbot should not be used to prepare a comment for the project’s issue tracker.
Project participants are additionally told not to discuss chatbot or LLM services in the covered spaces. The code of conduct describes those channels as places for focused technical work on MicroZig and other Zig Embedded Group projects, rather than forums for a wider debate about artificial intelligence. People who object to the policy are directed to contact maintainers privately or participate in other communities with different rules.
The translation provision is unusually broad. The project encourages English but does not require it. Contributors may write in their own language, while readers can rely on translation tools of their choice. What the policy rejects is contributors using an LLM to translate their own material before posting it to the project.
Enforcement is situated within the group’s broader moderation rules. The document emphasizes maintaining a constructive environment during technical disagreements and identifies maintainers Matthew Knight and Felix Queißner as contacts for reporting conduct problems. It says people who harass or antagonize others can be removed even when the conduct occurs outside the organization’s directly managed spaces.
The supplied document does not provide a detailed rationale for the LLM ban or describe a detection system. It therefore remains unclear how maintainers will distinguish machine-assisted material from unaided work, or how they will handle uncertain cases. The rule is framed as a participation condition: people joining the group’s channels are expected to follow it, regardless of whether other open-source projects permit similar tools.
That approach reflects a growing divergence among software communities. Some projects accept AI-assisted patches if contributors review and take responsibility for them. Others worry that generated submissions can increase review work, obscure authorship or introduce code that contributors do not fully understand. MicroZig’s policy avoids drawing a line based on quality or disclosure and instead excludes the use of LLMs across the contribution process.
The ban does not claim that every automated tool is prohibited, and the text supplied for this report specifically refers to LLMs and chatbot services. It also does not attempt to govern developers’ activity everywhere. Its stated scope is the Zig Embedded Group’s GitHub organization and Discord server.
For prospective contributors, the operational meaning is clear: code, documentation, issue comments and language assistance brought into those spaces must be produced without LLM involvement. The breadth of the rule makes MicroZig one of the more restrictive publicly documented examples of an open-source project setting boundaries around generative AI.
