The GCC steering committee adopted a policy on July 29, 2026 declining any legally significant contribution that includes or derives from LLM-generated content, using the GNU Project's roughly 15-line threshold for legal significance. LLM-generated test cases are the single carve-out, and using an LLM to research, find bugs or review patches stays permitted.
Read the full story: GCC Bars LLM-Generated Code Past the 15-Line Mark →
Transcript
GCC, the compiler behind most of the software running the internet, just decided how much AI generated code it will accept. The answer is basically none. The steering committee adopted a policy declining any legally significant contribution that includes or comes from an LLM, and they defined legally significant using a number that already existed in the GNU guidelines: about fifteen lines of code or text. Cross that line and it can't go in. You can still use an LLM for research, for finding bugs, for reviewing patches. You just can't ship its output. There's one exception. Maintainers can accept LLM generated test cases, because a test isn't the compiler, it just checks the compiler. GCC didn't invent a new rule here. It reused an old copyright definition and pointed it at a new problem, which is probably why this one sticks.