Thank you for saying this.
Other maintainers? Krita has only one maintainer.
Michelist
Ok, that’s clear enough. I’ll write up in general terms what I’ve figured out about what rhe feature is, what it requires etc. I doubt I’ll be able to implement it myself, it’d be a bit like writing a short story in French - I know the language a little but it would take me ages to get up to a level where I can get the task done.
After looking it up, it seems like there is one maintainer and a bunch of developers. With “other maintainers” I meant the other developers. Would need them to reject generative “ai” as well.
Really hoping they will, as this is an important issue. There is no ethical or responsible use of generative “ai”.
Halla, the maintainer, was pretty clear about not allowing LLM code, it would be pretty crummy (and unlikely) for a dev to try to sneak any in.
Maybe there could be a clear policy stated somewhere, so new contributors know? I tried looking before I asked but didn’t find any.
As one of the developers (and maintainer elsewhere), it is something I reject. I would also generally not be worried about the other developers involved. But I think most of us also don’t want to waste our time getting screamed at by techbros that neither use nor contribute to our software, so you probably won’t see us being too vocal about it.
That’s great to hear! 
It would be nice to have it written down somewhere, even if just in a CONTRIBUTING.md, that all usage of generative “ai” is forbidden. Everything from code completions to full vibe coding and everything non-code-related as well.
Just something like “It is forbidden to contribute any (partial or complete) outputs of generative models, this includes but is not limited to Claude, ChatGPT/Codex, Copilot, Gemini, Stable Diffusion, Midjourney, Qwen, DeepSeek, Mistral and Dall-E.”
It just being in a CONTRIBUTING.md would attract less attention, but still solidify it.
There’s been discussion and there may be something implemented in regards to this matter. But I neither want to kick off a(nother) discussion nor make wrong or empty claims, so that’s the extent of what I’ll say about it here.
Code-completion is not necessarily AI. If you program properly code completions is helpful. And everytime you, as programmer, have to decide if the completion is what you want. And it is local. So I don’t know what the problem with code completions is or should be. 
Code completion can use generative models, it depends what kind of completion. If this is just LSP completion, intellisense or such, then yeah, this is not using any generative model.
However, if this is something like Copilot, then this is using a generative model for completion.
Only the latter is problematic, the former has no problems.
Based on that, I recommend you ban any use of LLMs for coding, including a ban on LLM auto completion, as a public policy. Here’s what a ban would achieve beyond your statement here:
- It would make it clearer for contributors that aren’t reading the forums.
- It would reduce the amount of it happening. Case in point: Stealing code “manually”, e.g. by just copying code from somewhere else without the appropriate attribution, is already forbidden. You can’t enforce this 100%, but it’s still banned to make it socially discouraged and to make it clear that this is unwanted behavior that will get you kicked out if you do it. Banning LLMs works similarly, even if you can never 100% detect it if somebody tries to sneak it through.
- It would send a clear signal to other projects that banning AI is good, actually, and that you shouldn’t be afraid to do it despite the LLM hype. Here are other projects that banned AI: Asahi Linux, elementaryOS, Forgejo, Gedit, Gentoo, GIMP, GoToSocial, Löve2D, Loupe, NetBSD, postmarketOS, Qemu, RedoxOS, Servo, stb libraries, Zig.
- It would educate programmers on some code completion mechanisms being problematic, by calling out LLM code completion in the ban. People should know what technology they’re using and not blindly use all IDE features even when they’re harmful.
Here’s some more resources that could inform a ban:
- Clip of a lawyer seemingly doing a live demo of Co-Pilot in 2026. Literal quote: “This is a copyright infringement.”
- This study says the plagiarism rate of LLMs might at least be 2-10%, always, even when not trying to bait them into plagiarizing a known item:
https://dl.acm.org/doi/10.1145/3543507.3583199 - This study suggests that plagiarism is tied to model performance, so that the newer modern models might be even worse:
https://www.sciencedirect.com/science/article/pii/S2949719123000213Quote: “We found that the models that consistently output the highest-quality text are also the ones that have the highest memorization rate.” - Increasingly, new studies show LLM AIs don’t learn like humans at all and seem to be more like a lossy compression:
https://futurism.com/artificial-intelligence/ai-industry-recall-copyright-booksQuote: “Now, a damning new study could put AI companies on the defensive. In it, Stanford and Yale researchers found compelling evidence that AI models are actually copying all that data, not “learning” from it.” - This Apple study from 2025 suggests LLMs lack basic logical reasoning, which begs the question what these tools could ever possibly achieve other than plagiarizing with slight thoughtless adaptions:
https://www.forbes.com/sites/corneliawalther/2025/06/09/intelligence-illusion-what-apples-ai-study-reveals-about-reasoning/
I’m not an artist but I do use Krita occasionally (whether it be for silly joke images/memes or for temporary assets in my projects). I think a full ban on LLM tools would be a brilliant idea for reasons already discussed in this thread (and for the additional reason that its basically offloading our ability to think to a program that we have no idea how it truly works, which seems dubious and would probably cause a dependence issue)
To clarify, I joined this forum just to provide my 2 cents on this topic.
Speaking as someone who is neither a particularly great programmer (my sum total programming experience is beating PCem into submission to get a working if supremely janky port of it to macOS most of which was spent in gdb in iTerm2 and Stack Overflow on Firefox seeing why it failed to build or crashed) nor a visual artist (i mostly use Krita to do basic touchups to pre-existing images), I agree with the statements that LLM content as a whole should be outright forbidden from the project for basically all of the stated reasons so far, plus another one I haven’t seen yet: it also automatically makes it so any license you use for your project is null and void because non-humans cannot own copyright.
I wish more people outside of krita were as reasonable as the ones here.
Almost everything has been said here but I insist too. LLMs are not ethical no matter how you apply them to the code, a ban is the way to go.
Speaking as a software dev, a clear “no-AI” policy is likely to be the right move. Many pro-LLM-assisted-coding devs already don’t care much about the environmental and societal issues as long as they get their perceived convenience. However, there are plenty of additional reasons to avoid LLM-generated code in your project. These would be potential security vulnerabilities, licensing issues and the utter lack of quality control (and the resulting unperformant garbage cosplaying as “code”). A real productivity improvement is yet to be proven with data suggesting less productivity, not more. (https://hackaday.com/2025/07/11/measuring-the-impact-of-llms-on-experienced-developer-productivity/)
I think, you can share the code via your own fork first. Let people discuss more specifically. Maybe someone is willing to review it.
Also, I don’t know why you don’t try to understand the code with the help of AI explanations. You already know how to use AI.
Because GenAI technology is unethical from the start, that is why we object to it completely.
Ethical issues are way more important than anything practical.
Krita is free (as in freedom) software, which makes it an ethical alternative to the proprietary options. But now, with GenAI, there are more ways we have to make it ethical.
as a long-time user of krita, an official policy banning ai-generated code would be appreciated
Hi, not an artist just a longtime commissioner who has a great appreciation for art of all kinds, visual, musical, and code-icle. I know this is marked as solved now, but I’ll throw my hat into the ring and say that YES AI and LLMs should stay as far away from Krita as possible. Thank you to everyone who came together to oppose the idea and I’m sorry I’m late, but I want to add another name to the pile regardless. <3