RADV Vulkan driver adopts policy discouraging AI-generated code contributions

The RADV Vulkan driver for AMD Radeon GPUs on Linux has instituted its own policy governing contributions generated with large language models. Samuel Pitoiset, a Valve Linux graphics engineer and prolific RADV contributor, introduced the policy after observing an uptick in AI-generated patches from external contributors. The policy does not outright ban LLM-assisted contributions but discourages them, requires contributors to demonstrate good understanding of the codebase, and explicitly grants reviewers and maintainers authority to close such merge requests.

The policy also addresses LLM-generated bug reports, which are discouraged for being verbose and difficult to read, while permitting LLM-generated reproducers provided the contributor has personally reviewed them. Pitoiset stated the motivation was to establish clear expectations before the volume of AI-generated submissions became unmanageable for the small RADV maintainer team. The policy was merged into Mesa Git on November 2, 2026.

What the RADV LLM policy specifies

• Merge request handling: LLM-assisted merge requests are not banned but are discouraged; reviewers and maintainers may close them at their discretion if they lack sufficient code quality or author understanding.
• Issue reporting standards: LLM-written bug reports are discouraged because they tend toward excessive verbosity and reduced readability; reporters are asked to describe issues in their own words.
• Reproducer exceptions: LLM-generated reproducers are explicitly permitted as long as the contributor has personally reviewed them and confirmed they actually demonstrate the bug.
• Mesa layering: This RADV-specific policy exists alongside Mesa's broader project-wide AI/LLM policy, reflecting subsystem-level customization of contribution norms.
• Maintainer bandwidth rationale: The policy was initiated proactively due to observable growth in external LLM-generated patches, aimed at preventing reviewer overload before it escalates.

Why it matters

This development signals how specialized open-source graphics driver teams are adapting governance to manage AI-generated contribution flows without resorting to outright bans. For Linux gamers and Steam Deck users relying on RADV for AMD GPU Vulkan performance, preserving maintainer review capacity directly protects code quality and regression prevention. The policy may serve as a template for other Mesa drivers and kernel subsystems facing similar influxes of LLM-assisted submissions.