On March 17, Thomas de Grivel posted a ChatGPT‑generated ext4 implementation to the OpenBSD‑tech mailing list. The driver offers full read/write access and passes e2fsck, but omits journaling. De Grivel claims no Linux source was consulted, relying solely on LLM output and manual review. The code includes copyright assertions but no further disclosure.

Critics worry the LLM may have reproduced GPL‑licensed logic from the Linux ext4 codebase, risking license contamination. Christian Schulte highlighted that even documentation points to GPL material, making the contribution legally precarious. Theo de Raadt noted that reimplementation is legal, yet acceptance remains doubtful until a clear policy. OpenBSD team will consider it carefully.

OpenBSD’s lead developer, Theo de Raadt, emphasized that copyright status of AI‑generated code is undefined; without a clear holder, the project cannot redistribute it. De Grivel declined to retract, arguing AI can produce novel code without infringement, a stance that clashes with the project’s licensing strictures and safety guidelines.

The episode underscores a growing tension between open‑source projects and LLM‑driven contributions. As more developers turn to AI for rapid code generation, projects must weigh legal risk, maintainability, and author accountability. Until courts clarify AI copyright, OpenBSD will likely reject uncertain submissions and preserve project integrity.