rust-lang/rust adopts a living LLM contribution policy scoped to the compiler repo, erring toward restriction to simplify moderation of low-effort “slop” PRs.
Key Takeaways
Policy lives in Forge as a living document, linked from CONTRIBUTING.md and the rustc/std-dev guides; subtrees, submodules, and crates.io deps are out of scope.
Motivation is explicit: a deluge of low-effort LLM-generated PRs is overwhelming reviewers, and a standing policy beats case-by-case moderation.
Policy intentionally avoids a project-wide mandate; different teams (e.g., Clippy vs. compiler) may have different correctness standards.
A stricter full ban was rejected as likely to cause contributor exodus; the chosen middle ground bans usage that crosses a “threshold of originality” line.
Prior art section maps ~20 OSS projects on a spectrum from full ban (postmarketOS, Zig, Servo) to permissive (curl, Linux Foundation).
Hacker News Comment Review
Commenters highlighted the prior art spectrum as a standalone reference, useful for any project drafting its own AI contribution policy.
One commenter flagged a real scaling problem: LLM-generated Rust code can be featureful and complex enough that human reviewers lack bandwidth to evaluate it, creating a moderation bottleneck the policy itself does not solve.
