The Linux kernel just published official guidelines for using AI coding assistants.
It's a two-page doc. And it says more about where we're at than any hot take I've seen this week.
What it actually says
You can use AI tools to contribute to the kernel. But you own everything the AI writes.
Every line. Every bug. Every security flaw.
The Signed-off-by tag? Only humans can add that. AI agents are explicitly banned from signing off on commits.
Instead, there's a new tag: Assisted-by: AGENT_NAME:MODEL_VERSION.
If AI played a meaningful role in your code, you disclose it. That's the deal.
What Linus actually said
He doesn't want the documentation to become a "political battlefield" over AI.
His exact take: there's "zero point in talking about AI slop" in the docs, because bad actors who submit garbage AI code won't disclose it anyway.
The guidelines are for good actors. Everyone else is already a problem.
That's a pragmatic take you don't hear often.
Why the rest of us should care
Most of us aren't contributing to the Linux kernel. But the kernel's process is where software engineering norms get formalized first.
They invented the patch-based workflow. The DCO. The code review culture the entire open source ecosystem copied.
This is them saying: AI assistance is real, it's here, and we're going to treat it like any other tool — not ban it, not blindly embrace it, just hold contributors accountable for what they ship.
That accountability model is worth stealing.
The Assisted-by tag is a disclosure mechanism, not a judgment
It doesn't say "AI wrote this, be suspicious."
It says "a tool helped, here's which one, now the human owns it."
Compare that to how most companies handle AI-generated code right now.
No disclosure. No accountability. Just commits that look human until something breaks.
The Linux kernel just modeled what responsible AI contribution looks like.
Whether the rest of the industry follows is a different question.
Are you disclosing AI assistance in your commits? And do you think your team should?
👇
Top comments (0)