The flood gates for generative AI in Debian Linux are now seemingly open, with the results from a recent vote now available.
Generative AI is unfortunately pretty much everywhere in software development, and realistically - it will become impossible for any Linux distribution to outright ban it because it's even used in the Linux kernel now and across various other essential parts that make up Linux distros.
With the vote over, the Debian Linux team voted to allow "Responsible Use of Generative AI". Looking it up this is explained as:
🌐"
Using its power under Constitution section 4.1 (5), the project issues the following statement describing its current position on AI-assisted contributions. This statement describes the position of the project at the time it is adopted. That position may evolve as time passes without the need to resort to future general resolutions. The GR process remains available if the project needs a decision and cannot come to a consensus. Debian neither endorses nor prohibits the use of generative AI tools in the development, maintenance, or documentation of software, packaging, documentation, and other media published within the Debian Project. We recognize that such tools can substantially improve the productivity of contributors when used responsibly, allowing volunteers to spend more of their limited time on work that requires technical expertise, judgment, review, and collaboration.
The Debian Project nevertheless expects that all contributions submitted to Debian, regardless of how and with which tools they were produced, satisfy the same standards of quality, correctness, maintainability, and legal compliance. The use of a generative AI tool does not diminish the contributor's responsibility for the work they submit. Contributors are expected to understand, review, test, and, where appropriate, modify AI-assisted output before incorporating it into Debian. Blindly accepting or up AI-generated material without appropriate human review is inconsistent with Debian's established development practices. We enourage our contributors to disclose whether a contribution was made with AI assitance, but do not require them to do so.
Debian acknowledges that the legal status of material produced by generative AI systems remains the subject of ongoing discussion in many jurisdictions, including questions relating to copyright, authorship, licensing, and potential reproduction of training material. The Project does not seek to resolve these unsettled legal questions through this General Resolution, nor does it adopt a position on whether AI-generated output is, in whole or in part, copyrightable or derived from copyrighted works.
Instead, Debian continues to rely on the judgment and responsibility of its individual contributors. Project members are expected to exercise appropriate care when using generative AI tools, to consider the provenance and licensing implications of material they contribute, and to avoid introducing content whose legal status they cannot reasonably justify. Existing Debian policies governing licensing, copyright, software freedom, and the acceptance of contributions continue to apply irrespective of the tools used to produce those contributions.
Contributors are expected to exercise appropriate care when designing and implementing workflows that incorporate generative AI tools. In particular, they should ensure that confidential information, private communications, security-sensitive information (such as embargoed information about security bugs that is not yet public), cryptographic keys, credentials, and other non-public material relating to the Debian Project, its infrastructure, or its community are not disclosed to third-party AI services unless such disclosure has been explicitly authorized and is consistent with Debian's security and privacy requirements.
The use of generative AI does not alter Debian's established expectations regarding large-scale or automated project actions. Contributors intending to perform actions with broad project impact, such as mass bug filing or patch submission, large-scale code modifications, or other automated changes or requests affecting many packages or contributors, should seek prior discussion and consensus through the appropriate project channels before proceeding. Any such automated process should be overseen by a human who remains accountable for its behavior and output.
This resolution therefore affirms that generative AI is neither exempt from nor subject to special rules beyond the standards already expected of Debian contributors. The responsibility for every contribution rests with the contributor who submits it, who remains accountable for its technical quality, legal acceptability, and suitability for inclusion in Debian."
External Sources:
[lists.debian.org](https://lists.debian.org/debian-vote/2026/08/msg00360.html),
[debian.org](https://www.debian.org/vote/2026/vote_002#texte)
Some you may have missed, popular articles from the last month:
All posts need to
follow our rules. Please hit the Report Flag icon on any post that breaks the rules or contains illegal / harmful content. Readers can alsoemail usfor any issues or concerns.
On a completely unrelated note, I am going to get some additional backups of a core of files-of-high-importance. On immutable cold storage.
Last edited by emphy on 29 Aug 2026 at 9:39 am UTC
See more... I think it's interesting that these big FOSS projects tend towards the position that people are generally "good" and won't abuse genAI. It's refreshing, but it feels immensely naive to me. Perhaps we'll see updated policies when the reality of the situation kicks in with more and more Openclaw-like behaviours.
Just found this post on the thread by someone called Johannes Schauer Marin Rodrigues:
For myself this means that I rather "limit myself" and write software "slowly" or that I don't do some things at all. That way, I'm not part of the problems I see and instead am part of the world that I want to live in.That's a very enlightened outlook. Earlier in that post, he also acknowledges that it's a privileged position... but so is any position that is capable of paying £20 or £200/month for access to AI coding tool. Last edited by scaine on 29 Aug 2026 at 11:41 am UTC
See more... reallyirritating distro to use ime. But we're very quickly running out of ethical distros. Can there at least be a couple other distros that ban AI contributions? Even with the point of them not really being able to completely ban AI due to upstream projects accepting it, taking an ethical stance says and means a lot to anti-AI people like me. It at least means they're
to do something, unlike being lazy like Debian here.
trying The only other thing I can think of is something BSD based. NetBSD, like Gentoo, banned AI assisted/generated contributions, but that isn't really fit for desktop use. And the other modern BSDs haven't made any statements as far as I'm able to find.
I'm honestly just so disappointed. There has to be something we can go to that has some legitimately smart people behind it.
See more... Quoting: CurupiraAre you sure Proposal E won? Proposal H, which is critical of AI tools, has 17 votes.You're looking at proposals not at the voting results.
There were 8 options, four for AI (from extreme to moderate) and four against AI (from moderate to lenient), so ultimately the distribution was rather random. But when you sum op the actual options:
For:
Option 2 "Allow AI-Assisted Contributions with conditions" - 267
Option 4 "Accept AI contributions for Debian specific work" - 259
Option 5 "Responsible Use of Generative AI" - 281
Option 6 "A cautious approach to generative AI" - 276
Sum: 1083 Against:
Option 1 "Ban LLM contributions from Debian via Social Contract" - 144
Option 3 "Reject LLMs as far as practical, update Code of Conduct" - 176
Option 7 "Debian is created by humans" - 213
Option 8 "Avoid the use of LLM: climate destruction is a deal breaker" - 225
Sum: 758 So basically AI-enthusiasts outvoted AI-skeptics three-to-two, and every single pro-AI option had more votes than any anti-AI option. It's... disappointing, but not surprising at this point tbh.
See more... Quoting: scaineJust found this post on the thread by someone called Johannes Schauer Marin Rodrigues:The way I see it, and ignoring the social/environmental aspect, it might be sensible to use the AI for things that people generally hate doing, like code review (better to let it nitpick some bugs before they make it to production) or writing the documentation (better than reading 5000 lines of undocumented code trying to understand what its author had in mind), but it's absolutely insane to let it actually write code just for the sake of speeding things up(?). Solving problems and turning the solutions into code is the single thing that most programmers love, and here come people with their AI subscriptions and push "their" AI-generated code that nobody asked for... Like, man, if you don't like coding, just leave it to someone else, how hard is it to understand...?
For myself this means that I rather "limit myself" and write software "slowly" or that I don't do some things at all. That way, I'm not part of the problems I see and instead am part of the world that I want to live in.That's a very enlightened outlook. Earlier in that post, he also acknowledges that it's a privileged position... but so is any position that is capable of paying £20 or £200/month for access to AI coding tool.
[See more...](#)
I wish that people who respect the real expertise of software developers, especially in the linux development sphere, would also respect their same expertise to use Gen-AI as a lever for achieving contributions beyond their previous levels. Personally, I have built at work and at home projects of incredible scope (compared to my non-AI augmented work years ago). Because the machine generates code faster, yes partly, but mostly because it gives me access to such a large scope of domain expertise, that is both personalised and verifiable.
I have built software that span web development, networking, radiology, pathology, psychology, pharmacy, clinical research, embedded systems, I have installed whole computing environments from ordering the hardware to installing the software and networking it. I stand by every piece of work I produce, in some cases I published peer-reviewed articles on some of these projects. While some of my colleagues wait for a 30-minute meeting in 3 weeks for medical expertise from an expert that has trouble fitting their patients in their schedule, I can get the same answer in minutes to hours from Gen-AI because I know precisely how to cleanly validate an answer from a subject I knew barely anything about 5 minutes ago. In 2020 I did not know what a network port was and I had close to zero experience with python.
Now yes, using these tools has an environmental cost. However, I think everyone should try running Gen-AI tools on their own machines to really understand the costs involved. Since a lot of people here are gamers, most of them can actually run very capable models on GPUs they already own, while consuming electric power not greater than an average gaming session. When I run Qwen 3.8 27B on my 16 GB GPU at home, I am barely pushing 1.5 - 2x the power draw of a Playstation 5. And unlike a gaming session, a significant portion of that time is idle time because I have to actually damn think about the stuff the model is telling me.
I believe that the truly destructive and unsustainable practices come from the speculative hyperscaling behaviour of the big actors in the field that we all know. However, the Gen-AI tech that is already here, that doesn't need bottomless investments every year still, is already transformative, for better or for worse. And the better part of this transformation hasn't been realised yet, and we can get to it without spending billions in datacenters. We need smarter algorithms, smarter processes, smarter people, but we will also need Gen-AI. IMHO.
See more... I wish that people who respect the real expertise of software developers, especially in the linux development sphere, would also respect their same expertise to use Gen-AI as a lever for achieving contributions beyond their previous levels.Again, this disregards all the damage to society that is inherent to the tech.
Its very premise dehumanises, its foundational method of throwing more compute to the problem is inexcusably inefficient, and its output always will be cargo-cult quality at best.
See more... Quoting: emphyIts very premise dehumanises,This. AI dehumanizes and devalues everything it touches.
Equally concerning to me are the quasi-religious transhumanist views of its proponents, and AI's inevitable ubiquity in the world's militaries. AI is being used (or going to be used) to make tactical military decisions at the macro and micro levels, which is made possible by being trained on the bottomless amount of data that everybody feeds it. War itself has always been immoral, but something about giving a machine the autonomy to decide who gets to live and who dies just sends a shiver up my spine. We've always had the creativity and the tools to kill people in terrible ways - but I think we've really outdone ourselves here. "
It's just a tool", says Debian. "
It's useful", says Linus Torvalds. "
Can't beat em, join em", says Elon Musk. I'm thankful for the people that maintain the distros available to us, and for what great men like Linus Torvalds and Elon Musk have achieved - but I'll never agree with them here.
I choose not to use AI, but I often wonder why I bother because I know that this is futile. AI is something the vast majority of people in the world wants and supports because they've tasted what it's like. The illusion of "democratizing" skilled work by lowering the barrier of entry to creation is too attractive to the vast majority who believe that sacrificing the truth is a moral good, if it means that we get to live in a world where nobody is ever invalidated. There's no going back from this.
Quoting: MusangI see the polarisation caused by the use of Gen-AI, but as a software/AI practitioner I want to weigh in. While it may be naive to expect "responsible use" by all contributors, what is equally naive is the position that a large, complex software project can insulate itself from Gen-AI use in contributions.A very reasonable stance. I'm mostly in the same boat.
....
IMHO.
Won't make you many friends with the rather rabidly anti-AI crowd, but I assume you knew that coming in.
The actual Debian text is... awful, honestly.
While I'm fine with the sentiment behind it, it does not give anything concrete to go on. What exactly is or isn't allowed.
My impression is: anything goes.
And that is definitely a mistake, as it seems to allow AI-made commits.
That is really something that will likely become a large issue, as so many projects have shown, being drowned in low-quality PRs/MRs, draining manpower to review garbage.
If AI did or did not write the code in the end isn't relevant for a project (besides being borderline impossible to verify). What IS relevant is if the code is good or not, and aligns with the rest of the project in coding style, structure, etc.
This HAS to be ultimately reviewed by a human, and disallowing AI PRs/MRs (
including the entire description of it) is a simple way to filter out the most egregious garbage.
A way to try and make sure that the person responsible for the commit at least knows what they are doing.
This is IIRC what for example the Godot engine guideline is.
AI-assisted coding, yes.
Full AI-generated code + MR, no.
IMO they should also disallow commits by just anyone and vet people before allowing them to commit, but I don't know if they do that or not.
Last edited by TheSHEEEP on 30 Aug 2026 at 7:05 am UTC
See more... I honestly feel bad for kids starting their education now, and their teachers who have to ask themselves if the student wrote this, or some agent regurgitated a text from some Web site with a slightly different sentence structure...
See more... Quoting: emphyits output always will be cargo-cult quality at best.I understand many arguments against AI, and AI usage in coding, but this one is nonsense.
I have seen AI fail in coding, but I also have seen AI being incredibly good at it ("reasoning"/"agentic" AI, e.g. GitHub Copilot).
Which leads to another problem:
You/we will often be not able to tell if something has been coded with the help of AI.
"AI slop" is a thing, but some people seem to think that everything AI does is slop, and I can assure you it isn't.
Quoting: TheSHEEEPA very reasonable stance. I'm mostly in the same boat.Yes, the sentiment is admirable, some would say naive, but I agree that there should be an effort to put practical measures, even imperfect ones, in place to discourage irresponsible use. But I guess this is the difficulty of managing an inclusive community, you sometimes end up producing statements like this that will push users on both sides of the fence to more "opinionated" distributions...
Won't make you many friends with the rather rabidly anti-AI crowd, but I assume you knew that coming in.
The actual Debian text is... awful, honestly.
While I'm fine with the sentiment behind it, it does not give anything concrete to go on. What exactly is or isn't allowed.
My impression is: anything goes.
And that is definitely a mistake, as it seems to allow AI-made commits.
That is really something that will likely become a large issue, as so many projects have shown, being drowned in low-quality PRs/MRs, draining manpower to review garbage.
If AI did or did not write the code in the end isn't relevant for a project (besides being borderline impossible to verify). What IS relevant is if the code is good or not, and aligns with the rest of the project in coding style, structure, etc.
This HAS to be ultimately reviewed by a human, and disallowing AI PRs/MRs (includingthe entire description of it) is a simple way to filter out the most egregious garbage.
A way to try and make sure that the person responsible for the commit at least knows what they are doing.
This is IIRC what for example the Godot engine guideline is.
AI-assisted coding, yes.
Full AI-generated code + MR, no.
IMO they should also disallow commits by just anyone and vet people before allowing them to commit, but I don't know if they do that or not.
Quoting: EikeI understand many arguments against AI, and AI usage in coding, but this one is nonsense.I low-key think that managing Gen-AI answers is very much like building regular scientific method models in physics or biology. Level 0: the answer is sourced to an actual trustworthy website or content/equation that you can verify. Level 1: evaluate key statements of the answer by producing a falsifiable hypothesis from the statement. Run the falsification test to see if you can contradict it. Level 1+n: keep running falsification tests until one fails or you are satisfied with the odds of a mistake. I find this works identically whether you are testing generated code or medical explanations...
I have seen AI fail in coding, but I also have seen AI being incredibly good at it ("reasoning"/"agentic" AI, e.g. GitHub Copilot).
Which leads to another problem:
You/we will often be not able to tell if something has been coded with the help of AI.
"AI slop" is a thing, but some people seem to think that everything AI does is slop, and I can assure you it isn't.
Last edited by Musang on 30 Aug 2026 at 2:40 pm UTC
Quoting: MusangI have built software that span web development, networking, radiology, pathology, psychology, pharmacy, clinical research, embedded systemsThat sounds like a catastrophe waiting to happen.
Quoting: pbWait and see I guess... Anecdotally, the doctors who tested the output for a couple of years against gold standard routine and the reviewers who accepted the work found it acceptable. I hope the reviewers are like me and do not delegate the peer-review to Gen-AI... 😆Quoting: MusangI have built software that span web development, networking, radiology, pathology, psychology, pharmacy, clinical research, embedded systemsThat sounds like a catastrophe waiting to happen.