The Mesa RADV driver, which provides Vulkan support for AMD Radeon GPUs on Linux, has adopted a policy covering contributions made with artificial intelligence and large language models (LLMs). The initiative was led by Samuel Pitoiset of Valve’s Linux graphics team, a major contributor to RADV.
The policy does not prohibit AI-assisted merge requests outright. However, it states that such submissions require a good understanding of the codebase and impose a real review cost. RADV reviewers and maintainers may therefore close LLM-assisted merge requests at their discretion.
The project also discourages issues written by LLMs because they can be verbose and difficult to read. Contributors are asked to describe problems in their own words. AI-generated reproducers are allowed when they have been reviewed by the contributor and demonstrate the reported bug.
Pitoiset said the policy was introduced after the project began receiving more patches generated with AI from external contributors. He said it was preferable to clarify the position before maintainers became overwhelmed by such submissions.
The RADV-specific policy was merged into the Mesa Git repository today.
Sources: Phoronix.