GCC bans AI-generated code from its compiler, with exceptions for tests
The governing committee for GCC has adopted rules barring copyright-significant code from LLMs, even after human editing. Contributors may still use AI for research and isolated test cases.
The steering committee for the GNU Compiler Collection (GCC) has formally established a policy restricting the contribution of code produced by large language models, as reported by habr via Linuxiac.
Under the new rules, the project will reject any incoming contributions containing material generated by or derived from LLMs. This restriction extends beyond raw copy-pastes from assistants like ChatGPT, Gemini, or GitHub Copilot; it also applies to AI-generated code that has subsequently been edited or massaged by a human developer.
The boundary revolves around copyright significance. Contributors remain free to query models to explore concepts, decipher unfamiliar sections of code, or conduct background research. However, once content produced by a generative system reaches a threshold considered substantial under copyright standards, it cannot be committed to the core codebase.
The policy preserves two narrow practical carve-outs:
- Maintainers may accept legally insignificant AI-generated edits, provided they meet standard review criteria and the involvement of AI is explicitly declared.
- Test cases generated by AI are explicitly permitted. Because these are typically small, self-contained programs designed to reproduce compiler bugs or verify isolated behaviors, the project treats them separately from the executable compiler source itself.
Comments
Help shape the next version: Add context or suggest a correction. AI review can add points toward a rewrite. Reviews and updates may take time; a full meter does not guarantee a new version.