Fedora Linux now permits AI assisted contributions (code, docs and more), provided there is proper disclosure and transparency

https://pagure.io/Fedora-Council/tickets/issue/542

https://discussion.fedoraproject.org/t/council-policy-proposal-policy-on-ai-assisted-contributions/165092/242

Bad decision Fedora but it was expected when IBM/RedHat is basically fund it. Maybe that is what corporate client of RHEL/IBM wants. IDK. There are many other distros that bans AI/LLM but many contributions of Fedora ends up in upstream projects like Kernel or DEs.

Issue #542: Council Policy Proposal: Policy on AI-Assisted Contributions - tickets - Pagure.io

Human code reviewers of Fedora project can only address questions of quality, fix security issues, or address missing functionality in AI code slop. It cannot answer questions about legality or licensing or even trace origin like:

> Where did that AI code come from?
> What licensing does AI created slop follows? Can you put it under GPL?
> So who actually wrote the code? Why person submitting putting their name on top of it?

The answer is: Nobody knows.

My next guess is we'll hear about the Fedora Project reducing its developers because now AI slop can contribute to and maintain the distro. It's about cutting funding from IBM/Red Hat.

I am not a lawyer, so maybe they consulted IP lawyers or, worse, decided to just follow Microsoft's, Google's, OpenAI's, and others' idea of ignoring all copyright and ethical issues around it.

Interesting times are ahead and how the community will react it.

@nixCraft honestly I've been running Fedora as daily drive all 2025 and it has been good so far, it makes me mad that I am also trying to learn more about Ansible, which is under RHEL/IBM umbrella as well...not to mention that I was into Golang not so long ago, AI is circling everything I like and it feels less and less enjoyable every day.
@melocopon true. I lost all interest in IT field because of AI. It is just not enjoyable anymore.

@nixCraft there is an even worse: perhaps they decided that if their ai boat is made to sink by the courts, they can use it to drag down the open source boat with it.

If they can couple the legality of ai to the legality of opensource, they can force the opensource community to side with them. And if they lose they may be well positioned to benefit from what the opensource world loses.

@wdeborger @nixCraft I don't think so, since big companies benefit a lot from open source they would be shooting themselves in the foot. The only thing that big companies don't like is copyleft, and I don't think they can specifically target that. Open source in general just saves money, because people write code for free that those companies use themselves.
@nixCraft I don't expect AI-assisted contributions to Fedora to be a significant proportion anytime soon. Firstly, because most of the community doesn't want to use it, and secondly, the AI tech isn't good at what Fedora folks do.

@nixCraft My guess is that they just realize the inevitability that a non-zero amount of developers have features like Copilot enabled.

Banning any code-assistance is naive. It's like banning autocorrect or spell-check on documentation, there's no way to know it's not happening at all, and realistically it definitely is to some extend.

Better than to have a policy, so we can at least regulate it. Unless they're starting to allow literal bots to do MR's and be responsible, I'm not too worried.

@nixCraft

I think these are pretty wild guesses. AI can help people understand syntax errors, unknown codebases etc.

Fedora is a very complex project that will always need many developers. Note that most Fedora devs already do that in their free time, even though they might work their daytime job at Redhat.

The issues of licensing, stealing code and heating up the climate are severe and should be solved.

@nixCraft It really is a shame and reason enough for me, not to go back to using fedora on my workstation.

Weird, how every corp. now tries to "make AI work" when there's enough evidence that "AI everywhere" is just a very very bad idea. :/

@nixCraft Oh ffs... πŸ˜‘
@catsalad @nixCraft It doesn't actually look too bad, as long as everyone understands that, if you lack the ability to write the code yourself, you lack the ability to "[vouch] for the quality, license compliance, and utility" of the code. And if you *do* have that ability then, 999 times out of a thousand (at least), it would be easier to write it yourself than to check "the quality, license compliance, and utility" of the code after the fact.

@StarkRG @catsalad @nixCraft Ignoring all the ethical issues of AI generated anything (and we really really really shouldn't ignore those), my experience with AI is that it's like having to review code written by a junior that is incapable of learning.

Because that is my favourite part of coding... sure...

@StarkRG @catsalad @nixCraft And not even a talented junior, but one that got into coding without actually caring much about the craft.

@ainmosni @StarkRG @catsalad @nixCraft

Ladies and Gentlemen! This, and _exactly this_ is why the U.S.' "everyone should code!!!!1" push was so fucking stupid.

@desantis @ainmosni @catsalad @nixCraft
I disagree. I don't think that everyone should end up coding in their everyday life, however everyone should have a basic understanding of what code actually does. Without that you end up with people who think that computers are unknowable magic. You don't need to know how to genetically engineer something, but you do need to know what genetics *is*
@StarkRG @desantis @catsalad @nixCraft I think this learnt helplessness that we have with modern technology is a very big risk that has been exploited by big corps for quite a while now.
@ainmosni @StarkRG @catsalad @nixCraft Agreed, like how Cloud providers convinced the masses that you're not able to manage your own infrastructure.

@ainmosni @StarkRG @desantis @catsalad @nixCraft One of the many things I hate about LLMs is that this is essentially the opposite of software engineering, and yet so many software engineers seem to go along with it.

I'd hope that if more people knew the basics of coding, they'd recognize the problem. But that fact suggests otherwise.

It's as if medical schools were going all in on faith healing.

@foolishowl @ainmosni @StarkRG @desantis @catsalad @nixCraft
Well, medical profesionnals and WHO fucked up big time on covid and never updated their practices and recommendations (face respirators, clean air, etc.) sooooo in fact I am not sure that they’re better than software engineers. πŸ₯²
@StarkRG @desantis @ainmosni @catsalad @nixCraft Exactly. I shouldn't need to compile my corn from source, but I should be able to understand that it's fancy & drought resistant because of genetics.
@desantis @ainmosni @StarkRG @catsalad @nixCraft Everyone should code. Everyone should understand the computer. This LLM stuff is more like "no one should code or understand the computer".
@ainmosni @catsalad @nixCraft
The underlying architecture (deep learning transformers) has a lot of promise, LLMs, though, are absolutely pointless (and that's what all the non-image generative "AIs" are). On a scale between a Fisher Price kitchen set to a major commercial kitchen, LLMs rank quite a bit below easy-bake ovens. And then, as you mention, you have to factor in ethics and environmental impacts.

@StarkRG @catsalad @nixCraft Oh, there's promise, but as many of the AI researchers I've met have decried, LLMs have taken the wind out of the sails of the actual useful stuff.

I would have loved to see what would have happened if we weren't living under capitalism, somehow I think we'd have a lot less of this "can do the work of a human, if you drink 2 full litres of grain spirit before inspecting the work" nonsense.

@ainmosni As I've said to other people, I won't rule out deep learning transformers from being able to become a sapient entity (because they really are just that cool), but its training dataset would have to consist of *actual* *sapient* *thoughts* not just with written or spoken words. The only way something trained with words could appear to be a sapient entity would be if the observer has an incredibly deficient sense of empathy and theory of mind.
@catsalad @nixCraft FFS indeed! Just when I get settled on a β€˜it just works’ distro and it does this… πŸ€¦πŸ»β€β™‚οΈ

@nixCraft Time to watch where this contagion winds up.

Sometimes, the only way to deal with a pernicious disease is to amputate!

@nixCraft i think it is not a bad decision. World has change - it is a tool for contributors.
@nixCraft Desperately trying to find outlets for the monster.
@nixCraft shame on yous, fedora
@nixCraft That'll cause serious problems. DoD uses RHEL extensively, and also have strict rules about "no AI". I suspect Alma Linux is going to win bigly here.
@agreeable_landfall @nixCraft Yeah, but they also still use Windows 95.

@nixCraft It's really gross that they would decide to allow it, but I think they're putting the best possible restrictions on it by requiring the submitter to understand the generated code, but at that point why allow it? Why not just write good code yourself instead of rolling dice for it?

If this ends up causing problems (it will), does anyone have suggestions for what distro I should migrate my server to? Not Ubuntu or Arch, systemd preferred. I know Debian is good, but I don't like the release cycle; it's still my first choice despite that.

Void and Hyperbola are independent but neither use systemd. There's also openSUSE
@tinfoilchefspickaxe @nixCraft OpenSUSE apparently uses code from Fedora, so I would worry about that. What do you think of Hyperbola or Void for a server? I know Void is good for a desktop.
I'm not sure, never used Void myself. It has a version that comes without a desktop environment if you want that on a server

@jackemled @nixCraft im more worried about the ethical implications of allowing genAI explicitly, thus of genAI (ecological worries, digital labour, legal risks etc), i would think they *probably* have good quality assurance. but as you say, why bother then as most genAI contributions are of poor quality

if youre looking for alternatives to Fedora, and assuming you would prefer Fedora over other options due to, for example, quality and release cadence... then i would think of FreeBSD and Gentoo (even if im aware of the onboarding of the latter, it seems to be very good in regards to the control you have). i imagine declarative options like NixOS and Guix SD can also be considered, but Guix seemingly has quite old packages. personally im on Debian for most systems and likely going full Debian to replace Fedora, but you surely know enough to have gauged why it wasnt good for you

openSUSE could also be an option but i very much doubt it being a good alternative for our concerns, since you go from one distro whose corporate influence likely played in AI adoption towards another distro where... the corporate influence could lead into AI adoption just as easily

@arni @nixCraft I'll have to look at FreeBSD. NixOS would be perfect if the project management wasn't full of anti LGBT people.
@jackemled ah! well thats a thing i didnt know about 😭, thank you for the very fun fact
i thought of it some more after the fact and while its not systemd, Void Linux is very nice and could be worth checking out
@nixCraft time to change distro ?
I was about to reinstall Linux from scratch on my laptop (due to removing dual boot...) and thought going on with Fedora, but this news will make me think twice.
Now is time to find another distro that will fit my needs...
@jfpoilpret @nixCraft It might be worth waiting a few months to see how Fedora turns out. I dislike that they're allowing contributions generated by LLMs, but it seems like they're going to require those contributions to be labeled & like they're going to be held to a high standard, which is probably the best way to handle the situation. It could be that they were being flooded by crappy contributions by ChatGPT or something, & they thought this might be a good way to get vibe coders to out themselves so someone responsible can either ban them or give their code a very strict review.
@jackemled @nixCraft actually, I could not wait that long, I finally installed Debian 13 today, just removed both Fedora and Windows which I had kept for too long in one partition.
Now I'll need some more time to install everything I need.
But I am already happy with this first step.
It didn't go as well as I expected: issues with BIOS, with GRUB... All finally solved after some cold sweat πŸ˜…
@jfpoilpret @nixCraft I hope the rest of the setup goes well! I might try out plain Debian too sometime.
@jackemled @nixCraft I chose Debian for the sake of consistency, as I use it already on a VPS, that should reduce my admin time (no need to try to remember 2 ways of doing the same admin tasks).
After my notebook, I'll still have my PC to reinstall from Fedora to Debian, then I plan to soon replace my VPS which will be also on Debian.
I love consistency!
@nixCraft The Fedora update system delegates trust to the mirror sites, so its part of a pattern of making RHEL's relatives appear unstable and untrustworthy.
@nixCraft guess its a good coincidence that i thought of going to debian for my home pc, or gentoo if im feeling brave enough 🫣