Rust Project Forbids AI From Creating Code, Permits It to Suggest 1,200 Consecutive Lines; Working Group to Clarify Difference

INTERNET— The Rust programming language project this week published a policy permitting contributors to use AI language models to "answer questions, analyze, distill, refine, check, suggest, and review" code while explicitly forbidding models from "creating" it. The document does not define creating.

By Wednesday, maintainers had received 14 compliant disclosure statements. Three described the contributor's model as having "suggested" between 800 and 1,200 consecutive lines of code, which the contributor then "accepted" after, in each case, "light review." Maintainers asked whether this constituted creation. Contributors said it did not. The thread remains open.

"Reviewer bandwidth is finite," the policy notes. "LLM-generated code places additional burdens on reviewers, whose time is a precious community resource." Since publication, the project's open PR queue has grown from 2,847 to 3,409 items, as contributors sought advance rulings on whether their planned AI usage would qualify as analysis or authorship. No rulings have been issued. A working group will convene when quorum is available.

One proposed enforcement test — whether the AI "initiated" a code block or merely "completed" a human-initiated one — generated 47 replies before stalling on the question of what initiation means.

A second proposal, distinguishing AI outputs that "implement a concept" from those that "refine an existing one," was described by three maintainers as "promising" and by a fourth as "the same problem."

The Rust project confirmed the policy itself was drafted with the assistance of an AI language model, which was asked to "distill" months of maintainer discussion into coherent language.

Whether distillation constitutes creation has not been determined.