What Rust’s New LLM Usage Policy Means for Contributors

Policy Overview

Rust team released a policy document governing the use of large language models (LLMs) in the rust-lang/rust repositories. The document defines permitted and prohibited actions.

Origin and Publication

The policy originated from a Zulip discussion thread containing over 3,000 messages. It is stored on the Forge platform as a living document, linked from CONTRIBUTING.md and the rustc and std development guides. Contributors are expected to check the document regularly because the rules may evolve.

Scope Exclusions

The policy explicitly excludes broader topics such as long‑term societal and economic impact, environmental impact, copyright status of LLM‑generated output, and moral judgments of LLM users.

Motivation

Reviewers observed a surge of low‑quality, AI‑generated pull requests that made code review harder. The policy aims to simplify review by providing clear boundaries instead of handling each case individually.

Prohibited Uses

The policy bans certain LLM‑specific syntax and imposes strict limits on AI‑generated code or text within the compiler or standard library work. The approach favors restriction over outright prohibition.

Comparison with Other Projects

During drafting, the team examined precedents:

Projects that completely forbid AI‑generated contributions: postmarketOS, Zig, Servo, QEMU.

Projects that allow AI assistance but require human supervision: SciPy, LLVM, Blender, Linux kernel, Firefox, Ghostty, Fedora, curl, Linux Foundation.

Rust adopts a middle path: AI tools are not banned, but contributors must take responsibility for any code they submit.

Feedback Guidance

The authors note that some sections may be uncomfortable and ask developers to consider whether a proposed change is easier to manage, the cost of not having a policy, workflow impact, and tolerance for interruptions.

Conclusion

The policy permits AI for learning, problem discovery, or review assistance, but requires that actual code contributions be made without AI assistance, reflecting a focus on code quality and community standards.

Reference: https://github.com/rust-lang/rust-forge/pull/1040