Mesa's RADV Vulkan driver adopts its own policy on AI-written code

The RADV Vulkan driver, which provides Vulkan API support for AMD Radeon GPUs on Linux as part of the Mesa graphics project, has adopted its own policy governing AI and large-language-model use in contributions, on top of Mesa's existing broader AI/LLM policy. Samuel Pitoiset, a Valve Linux graphics team member and prolific RADV contributor, led the effort to add the driver-specific rules.

The policy does not ban AI-assisted contributions outright, but discourages them and gives maintainers room to close merge requests that were AI-generated at their discretion. The text states that LLM-assisted merge requests require a good understanding of the codebase and that reviewing them carries a real cost for maintainers. It also discourages LLM-written issue reports, saying they tend to be verbose and hard to read, and asks contributors to describe issues in their own words instead; LLM-generated bug reproducers are allowed as long as the contributor reviewed them and confirmed they demonstrate the bug.

Pitoiset said the move was meant to get ahead of a rising number of AI-generated merge requests from external contributors before maintainers become overwhelmed, writing in the merge request that the project was starting to see slightly more LLM-based submissions and it seemed better to set expectations now. The policy was merged into Mesa's Git repository.

Why it matters

As AI-assisted contributions grow, open-source projects are starting to set explicit ground rules to manage reviewer workload rather than banning the tools outright.

Source: Phoronix