rust-lang/rust adopts an LLM policy that permits review and analysis but restricts creation

Sources: primary, discussion
Summary: Jynn Nelson writes that five teams adopted a policy governing LLM use in the rust-lang/rust monorepo, and states it is not an official Rust project stance and does not apply elsewhere in the project. The policy summarizes itself as fine to use LLMs to answer, analyze, distill, refine, check, suggest and review, but not to create. Disclosure is required for machine translation, trivial changes, discovering bugs, and reviewing others' work with an LLM, and unmarked LLM output is not allowed in public docs, PR descriptions or GitHub comments. LLM-generated code changes must be pre-arranged, non-critical, well-tested and disclosed, are required to have tests regardless of difficulty, and must not touch soundness-critical areas unless the author is already a domain expert. Reviewers may close non-conforming PRs without discussion and are never required to review LLM PRs. The post cites 1,281 open PRs as the review-bandwidth pressure behind it and states that harassing someone for LLM use is itself a violation.
Why it matters: Rust holds LLM-authored changes to a higher bar than human-authored ones, which inverts the usual framing, and it is a different rule from the drivers/staging rejection carried in Top stories rather than the same trend.