For years, the Linux kernel community has required patches developed with AI/LLM assistance to carry an Assisted-by tag. The goal was to make the partially artificial origin of the code explicit, giving maintainers and reviewers more context. Now developers are weighing the removal of that requirement, arguing that the pervasive use of LLMs makes the tag impractical.

The proposal is not just a bureaucratic simplification. It reflects a structural shift in how software is produced: when LLM assistance is the default, distinguishing human from synthetic contribution loses informational value. In a development flow where generative models are embedded in editors, autocompletion tools, and review pipelines, a single tag risks capturing only a fraction of the support received. The rule becomes hard to apply consistently.

Who loses from this change? Anyone who needs to audit the kernel's supply chain. Provenance tags like Assisted-by offer a low-cost clue about the nature of a modification. Without them, responsibility shifts to more invasive analysis, such as deep manual review or pattern-monitoring systems. Organizations running the kernel in regulated environments, or managing self-hosted infrastructure for data sovereignty reasons, may end up with less information for risk assessment. The audit cost does not disappear; it moves from the kernel community to individual organizations.

At the same time, the move signals normalization of LLM use in development. The question is no longer whether an LLM was involved, but whether the resulting code is correct, maintainable, and secure. This shifts attention from the producer to the product. It is, in a sense, an acknowledgment that technical review must rest on the code itself rather than the story of how it was written.

But that approach has a problem: kernel reviewers already work with limited resources. If the tag disappears, they lose a signal that could direct attention to synthetically generated modifications. Not all assisted contributions are lower quality, but knowing an LLM played a role helps reviewers look for specific error patterns common in generative models. Removing the tag does not eliminate those risks; it makes them more opaque.

The Linux kernel community is used to pragmatic governance decisions. But dropping the requirement to disclose LLM use comes at a time when supply chain transparency has become a priority for governments and companies. The open question is whether automated detection tools will fill the gap left by the tag, or whether the ecosystem will simply accept reduced traceability. For those evaluating on-premise deployments, the issue is concrete: the kernel is the foundation on which trust in infrastructure is built, and the trade-offs between software transparency and development speed are not secondary. AI-RADAR explores these topics with analytical frameworks at /llm-onpremise.