Flathub, one of the most popular ways to grab applications on Linux, has a newly updated generative AI policy - where it's pretty much all banned. However, there is an exception noted for "mature, well-maintained projects" but it's not a guarantee.
A new commit was sent in and merged into the documentation, which is live now, that notes "Reword LLM policy to make it clear it's not allowed". The new policy reads:
Generative AI policy
This policy applies to both the application being submitted to Flathub and the Flathub submission itself, including the manifest, metadata, patches, build scripts, and pull request. For the purpose of this policy, applications include BaseApps, extensions, and any other artifacts that can be produced by flatpak-builder.
Submission pull requests must not be generated, opened, or automated using AI tools or agents. Please also do not request review from any AI tools in the submission PR. Automated Copilot reviews on GitHub can be disabled by the submitter by going here and changing
Repository accessto exclude the repo or disabling the global "Automatic Copilot code review" found here.Applications containing AI-generated or AI-assisted code, documentation, or other content are not allowed.
Applications or changes containing copyrighted, license-incompatible, or ethically questionable code are not allowed.
These submissions can be rejected without any further review.
Repeatedly violating these policies may result in a permanent ban from future submissions and activities.
Exceptions may be granted for mature, well-maintained projects.
To give some more context, developer Bart Piotrowski mentioned in a social media post on Mastodon:
We have updated Flathub's LLM policy to explicitly disallow AI usage for both the submission process and applications being submitted.
https://github.com/flathub-infra/documentation/commit/992f57b30de98ddbd5e80959e9672998c83c8c97
I've had some reservations about it, so the wording before that commit was relatively milder. I know it's an unpopular opinion on the Fediverse, but I do think LLMs are inevitable, and the reality is that you can expect less organically grown code as time goes on. I believe it can be a useful tool in and outside FOSS; I hoped we will see a larger number of apps where authors made some effort beyond prompting an agent. Meanwhile, the number of unpleasant interactions I've had with entitled submitters acting as if they were bestowing their brilliant software upon us idiots who are rejecting it went through the roof in the last month. I'm tired.
As always, we are not applying this retroactively, so any vibecoded apps which were already published will remain available.
What are your thoughts on this? No matter which side of the argument you're on, having clearly defined rules around it is a good thing so that it's clear for everyone.
Article taken from GamingOnLinux.com.
See more from me
You can also find comments for this article on social media: 
Quoting: CatKillerin my opinion, they mean: everything that was stolen from GPL or proprietary software shouldnt be used .Applications or changes containing copyrighted, license-incompatible, or ethically questionable code are not allowed.They're going to have to clean up the wording on this: all code is copyrighted.
permissive licences like mit, might be used.
spacemonkey 19 hours ago
This.
Last edited by tohur on 29 May 2026 at 5:11 pm UTC
Johnologue 18 hours ago
I know it's an unpopular opinion on the Fediverse, but I do think LLMs are inevitable, and the reality is that you can expect less organically grown code as time goes on. I believe it can be a useful tool in and outside FOSS; I hoped we will see a larger number of apps where authors made some effort beyond prompting an agent. Meanwhile, the number of unpleasant interactions I've had with entitled submitters acting as if they were bestowing their brilliant software upon us idiots who are rejecting it went through the roof in the last month. I'm tired.I think "LLMs are inevitable" is a terrible phrasing for a lot of reasons (it's like the "communism is great on paper, but..." thing - not about the stance, it's that people repeat this generic phrasing that doesn't actually make sense!), but I basically agree.
My current leaning is that it seems like LLM-based coding systems have a useful future. The problem is social/human. Corporations using it to gain power in extremely harmful ways...and individuals who have decided that simply using AI makes them superior to people with genuine skill/practice. They're more an extension of "AI artists" than they are "AI coders".
It is sensible and necessary to filter out AI submissions as a whole at this time.
I suppose this is also a good opportunity for Flatpak as a format. If an AI-based project distinguishes itself, people don't need to wait for Flathub to package/access it as a Flatpak, because they can be distributed outside of that centralized "app store". It raises a social barrier, but one that can be overcome.
Arthur Prazeres 18 hours ago
Quoting: tohurTo be honest all this is going to do is make people not be honest about using AI.. you people want transparency but the route your taking will only hide the fact people are using AI because 1000% bet you people can NOT tell human code from properly iterated and tested AI code, and now these coders because I again 1000% the people using AI properly are actual coders to be able to get the AI to produce code you won't be able to tell if its human or notyou have defensive ai tools for that, like pangram and winston
Draconicrose 17 hours ago
The contents of a repo are what makes or breaks it.
Some repos reject all closed source stuff, Apple rejects what it can't control, Many reject malware, github rejects nudifying AI, debian14 rejects non-reproducable packages
Many of these policies are perfectly reasonable and sometimes lawfully enforced, but they're all an integral part of the repo experience.
I mean, maybe he wants to keep doubling down on this, he certainly can. But I think some major projects are going to start pulling their work in protest (KDE for example?)
But the current way it is written is way too restrictive.
This is essentially going to kill Flathub if it remains like that.
Mostly because of this part:
Applications containing AI-generated or AI-assisted code, documentation, or other content are not allowed.The vast majority of developers use at least AI-assistance in software development nowadays. A few years from now, you basically won't find anyone anymore not doing so (except a few luddites, I guess, but they'll find fewer and fewer gigs).
Besides being completely unable to verify that to begin with.
For example, if I use AI to answer me a bunch of questions about the code base in front of me, then make changes in the code on my own, that would still be AI-assisted.
Same thing if you want to switch some file to a different coding guideline, but let AI do it. Result would be identical to what anyone would do manually, but suddenly it isn't kosher anymore.
Or if you changed 100 lines of code, AI detected a typo and fixed a line for you.
Or even if you ran it on some code to detect issues and fixed those yourself.
The amounts of legitimate use cases for AI, without letting AI do the entire coding work and produce garbage, is huge.
Scratch the AI-assisted part of that requirement and I think it would be quite reasonable.
And this part, too, is just strongly eyebrow-raising:
These submissions can be rejected without any further review.So someone as much as believes something was even just AI-assisted, bam! - that's it. 🤣
Quoting: Arthur PrazeresYou have defensive ai tools for that, like pangram and winstonThey can of course detect some blatantly obvious stuff like entire commits & pull requests, etc. being made with AI.
But you cannot detect assistance, as it may not even lead to any code being written by AI at all.
Last edited by TheSHEEEP on 30 May 2026 at 10:39 am UTC
The amount to sift through on public repositories is huge, and that is generally done by a human - who would then spend significant amounts of time looking at garbage code "produced" by some vibers.
Understandable that that's not a job anyone would want to do willingly, many probably not even for money.
The best workaround I see is to use the defensive tools available, as mentioned previously.
Those can at least filter out the most obvious offenders, and maybe raise flags for some other stuff so humans can then look at it.
It'll still be more than it used to be prior to AI tools...
But yeah, ultimately this issue has not been resolved in any satisfactory way for projects that anyone can send pull requests to.
Maybe the only solution is to no longer allow pull requests from any rando and only allow actually vetted team members to send PRs.
And then maybe collect potential changes "from outside" elsewhere, also in a way that should only allow actual humans to suggest them there.
But how?
🤷♂️
Last edited by TheSHEEEP on 30 May 2026 at 10:33 am UTC
Patreon
PayPal
Anticheat check - which competitive games actually work on Linux?
How to give Valve feedback when Proton games have issues on Linux / SteamOS