Rust Sets New Rules for AI Coding: AI Can Help You Look, But Not Write for You; Overuse Triggers 'Circuit Breaker'

On August 5, 2026, Jynn Nelson announced on the Inside Rust blog that five teams within the Rust project—compiler, standard library, type system, rustdoc, and bootstrap—have formally adopted a policy governing the use of large language models (LLMs) in code contributions. The policy applies to the main rust-lang/rust repository and states that LLMs may be used for answering questions, analysis, summarization, refining, checking, suggesting, and reviewing, but not for creating new content. Contributions that involve LLM creation must be disclosed, and new contributors are not allowed to submit LLM-created pull requests unless they first find a reviewer. The policy includes a 'circuit breaker': if more than half of merged PRs in any six-week window are LLM-created, all such PRs will be paused until the proportion drops below 50% and a 10-day cooldown passes. Violations, such as deliberately concealing LLM use, are treated as violations of the Rust Code of Conduct, just like harassment. The policy is intentionally not fully enforceable technically, aiming to eliminate ambiguity rather than catch every violation. Nelson cited three pressures that led to the written policy: the signal of a well-crafted PR is no longer reliable, review resources are becoming scarcer as code generation costs drop, and some contributors copy reviewer comments into LLMs and paste outputs back, wasting everyone's time. The policy is seen as a first step, with the Rust leadership council considering a project-level LLM committee that could expand such rules across the entire project.