Zig's Rationale for Strict Anti-AI Contribution Ban
Zig maintains one of the toughest policies against large language models among significant open source efforts. The project forbids LLM-generated comments on its bug tracker. This includes any translations. The team encourages English but does not demand it. Contributors can write in their native tongue. Others may use preferred translation tools to understand the posts.
Prominent Zig Project and Its Fork
Bun stands as the best-known project built with Zig. This JavaScript runtime saw acquisition by Anthropic in December 2025. Not surprisingly, Bun relies extensively on AI tools. Bun runs its own version of Zig. Recently, it gained a 4x speedup in Bun compile times. Developers added parallel semantic analysis and multiple codegen units to the LLVM backend. The code appears here. But @bunjavascript stated: "We do not currently plan to upstream this, as Zig has a strict ban on LLM-authored contributions."
Bun's choice highlights tensions between AI use and Zig's rules. Zig, launched in 2015 by Andrew Kelley, serves as a systems programming language. It aims to replace C with better safety and simplicity features. The language emphasizes manual memory management and comptime execution. Bun, created by Jarred Sumner, competes with Node.js as a swift alternative for running JavaScript. Its speed comes partly from Zig's compiler strengths. Anthropic, an AI safety firm founded in 2021 by ex-OpenAI staff, bought Bun to boost its tech stack.
Loris Cro's Explanation of the Policy
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Zig Software Foundation VP of Community Loris Cro detailed the ban's reasoning. His post, "Contributor Poker and Zig's AI Ban," shared via Lobste.rs, offers the clearest case yet for rejecting all LLM-assisted work. In thriving open source projects, pull requests often outpace review capacity. Projects might reject flawed submissions to focus efforts. Zig takes a different path. The team aids newcomers to refine their submissions for inclusion.
This approach stems from more than goodwill. It proves practical too. Zig prioritizes people over code. Each contributor receives investment from the core team. The main aim of reviews goes beyond merging patches. It builds skilled, reliable participants for the long term.
How LLMs Undermine Contributor Growth
LLM help shatters this model. A perfect pull request from an AI tool changes nothing. Review hours fail to cultivate assured, dependable team members. Cro calls it "contributor poker." Like the card game, players focus on opponents, not hands. In this version, maintainers wager on individuals, not initial submissions.
The concept aligns with another view. If an LLM drafted most of a pull request, why have humans review it? Maintainers could prompt their own models for solutions. Simon Willison, a noted developer behind Datasette and a prolific blogger on tools like Python and LLMs, highlighted Cro's points. Willison posted this note on 30th April 2026.
Zig's stance reflects broader debates in open source. Projects face floods of AI-generated code. Maintainers weigh quality against mentorship. Zig bets on human growth. This policy shapes how contributors engage. It demands original effort. Bun's fork shows workarounds exist. Yet upstream paths stay closed to AI-touched code.