A lot of work and code is Debian specific, and so there's a choice to be made about LLM use for that code, as there is for any project. And there are questions about LLM use in mailing list discussions, bug reports, and other non code interactions. Most of the options take a position on that (i.e. impose some rules for those contributions) one way or another.

What other projects choose on the question is interesting and might inform the choice, but what e.g. the kernel opts for doesn't control what Debian decides for itself, any more than the kernel's choice imposes any requirements on Rust or glibc or gcc.

What is the point of setting standards on the way we work? It's an exercise in democratic decision making. Some people dislike ai for entirely legitimate reasons, and they want to see if others in their community agree. (And if so, set boundaries on the way they work.)

You gotta stop seeing "virtue signaling" every time someone holds a belief different than your own.

They care about it. And want there to be an official position about it.

It means that for any patches submitted directly to Debian, they can blanket reject it if it has signs of AI without having to properly evaluate the patch on it's merits (which is onerous and easily DDOSed).

What virtue would be signaled in this case and to whom?

While it would be desirable to ban all packages that use AI, Debian has its own code to protect.

They cannot ban all packages because Torvalds has been bought by the sponsors of the Linux Foundation and has already used the "AI is a tool" talking point from the corporate manuals.

Debian and Linux partly started as resistance projects against Microsoft, so a moral stance is expected and welcome anyway.

We'll see if the Canonical people who always vote for corporations or the idealists win. If Canonical wins this, Debian is obsolete.