Debian、LLM 利用の提案を審議。オープンソース開発者への影響

The board behind open source operating system Debian has tabled proposals for and against the use of LLM-assisted contributions in Debian, in a general resolution statement.

Basing its stance on Debian’s “reputation for stability”, a proposal attributed to Matthias Geiger stated that, as a community, the group feels that widespread LLM usage comes from the “move fast and break things” attitude.

It’s just not what makes Debian, Debian

“While [that approach] is common in many parts of this industry, it is contrary to what makes Debian Debian, and is inappropriate for Debian contributors. We will not allow direct contributions to Debian written with the use or assistance of large language models (LLMs) or other generative AI tools.”

Seven counterarguments are also used to table proposals that include: allowing AI-assisted contributions with conditions; rejecting LLMs as far as practical, while also updating the code of conduct; accepting AI contributions for Debian-specific work; guidelines on responsible use of generative AI; a statement written under the pledge “Debian is created by humans”; and the unceremoniously stark final proposal entitled “avoid the use of LLM: climate destruction is a deal breaker”, which states that planet earth is burning.

Named after its originator Ian Murdock as a portmanteau of his own name and that of his then-wife Deborah, Debian retains a strong fanbase among developers and is said to have replaced Windows when it was deployed on the International Space Station back in 2013.

Still only a set of proposed limitations, if approved by the committee, the outright ban stipulation would apply to Debian source packages and other software developed by Debian. It would also apply to official Debian web resources, direct code contributions defined as packaging, native Debian software such as the lintian package checker, as well as documentation and translations written by Debian contributors.

Any AI prohibition on Debian would not apply to upstream software engineering projects that make use of generative AI and, equally, it would not come into force to ban the use of patches and security fixes from upstream sources that may or may not have used AI services. While Debian developers could still find themselves packaging third-party AI-coded software tools, they would not be allowed to use those tools for new direct project work.

“Debian’s debate shows that it agrees that provenance must be declared, but this eight-way vote is an argument about how big the sticker should be on code origins. That is not a war over principle – that is a formatting dispute with a manifesto attached.”

An eight-way vote argument over a provenance sticker

Author of the Hybrid Workforce Standard, Joe Phillips, tells The New Stack that Debian’s debate here shows that it agrees that provenance must be declared, but this eight-way vote is an argument about “how big the sticker should be” on code origins.

“That is not a war over principle – that is a formatting dispute with a manifesto attached,” Phillips says. “The Debian community’s strongest argument is not copyright, and it is not the environment. It’s buried in the community section, which states: ‘contributors who lean on a model never learn packaging, so they can never replace the maintainer who burns out’ and that highlights deskilling, so it’s the one harm here that compounds.”

Philips says it’s the same risk factor he requires organizations to name and measure in his hybrid workforce consultancy and, he feels, a ban teaches nobody anything.

“If Debian cares about its pipeline, the answer is apprenticeship, not prohibition: make new contributors work in shadow mode, review them like juniors, and let them earn authority,” asserts Philips.

Right kind of idea, wrong fix for the problem

CEO & founder at AI-native compliance management platform company Strike Graph, Justin Beals, tells The New Stack that this move is “the wrong fix for the actual problem” at hand.

“Debian’s ban isn’t really about banning AI,” Beals says. “It’s an admission that nobody has built a reliable way to verify what an AI agent actually produced before it lands in a codebase this many systems depend on, including infrastructure running in orbit. That’s a legitimate thing to be worried about.”

He reminds us that every open source project running on trust and reputation is now exposed to contributors who can generate infinite plausible-looking code, and so move fast and break things was already the wrong model for infrastructure this critical, long before LLMs showed up. He thinks that a blanket ban doesn’t close that gap, it just removes one specific way of triggering it while leaving the underlying trust model untouched.

“Banning LLM contributions treats the symptom, not the disease,” Beals emphasizes. “In the wake of Debian’s AI-code knee-jerk, the projects that come out ahead here will be the ones that build verification into the contribution pipeline itself: provenance tracking, review depth tied to actual risk, and evidence that a human meaningfully validated what shipped, regardless of how it was written.”

He thinks that any policy statement that says AI isn’t allowed simply won’t hold up, as the tooling keeps getting harder to detect.

“Banning AI-assisted code doesn’t eliminate AI-assisted code – it’ll just make its use harder to disclose… what’s stopping someone from copy/pasting everything, or even manually typing out what it generated?”

It’s not AI-assisted code if I type it all out again

Principal AI architect at Endor Labs, Matt Brown, tells The New Stack that he understands the Debian team’s concerns. He feels a project built on stability can’t afford unreviewed, low-quality contributions simply because AI makes them faster to produce.

“But any blanket ban risks confusing the tool with the quality of the work, especially as these models become more capable and widely used,” Brown says. “Banning AI-assisted code doesn’t eliminate AI-assisted code – it’ll just make its use harder to disclose – what’s stopping someone from copy/pasting everything, or even manually typing out what it generated?. A better approach would require transparency, human accountability, rigorous testing, and the same high review standards for every contribution.”

For Brown, the real question is not whether AI was involved, but whether the contributor fully understands, verifies, and stands behind the code.

How Debian’s proposals might pass

Looking across the complete set of tabled propositions, proposal A (no LLM contributions to Debian via social contract) needs a 3:1 majority to pass; the other seven proposals (B to H) need a simple majority. Voting closes Friday, 2026-08-28 23:59:59 UTC.

The post Debian just proposed banning AI code. Here’s why it matters for open source developers & maintainers. appeared first on The New Stack.