many people are posting about Godot banning AI contributions right now, but there's a catch.
In their announcement, they say:
No use of AI to generate substantial pieces of code
(emphasis mine)
that is to say: you're still allowed to use it as long as the influence of AI in a PR is small enough.
which is an important distinction which makes me a bit sad, because it means it just about doesn't make the cut for the No-AI List
anyone here part of the Godot Foundation and wants to do a small push for covering this "last mile" too?
Jul 1, 2026, 9:16 AM
Jul 1, 2026, 9:16 AMfine, I'll be more specific.
for the record I think it's weird how they list 3 separate things that don't require AI as examples
Jul 1, 2026, 9:29 AMThey don't specify whether it's a language server or an AI tool. They just say code completions, which is too vague. If they mean only one of these two things, they should specify it. Right now, because they just say code completion, an LSP would still fall under this, because it's still a type of code completion. Of course, I'm sure that they evaluate this on a case-by-case basis, but their policy document should be clear about this.
@alexia@app.wafrn.net @risc@wetdry.world
The wording on the blog post isn't great, but The current not-yet-amended version of the policy already has a this disclaimer on it which should make it reasonably clear they don't mean LSP
"AI" in this page refers to any LLM/generative AI model like ChatGPT, Claude, Grok, etc., or any other models that fulfill the same role.I imagine the final amended policy will have a similar disclaimer, and by code completion they probably mean single-line completion like what Intellicode does (Which, unlike Intellisense, is AI-based and also enabled by default in Visual Studio. Thanks Microsoft), which is also specifically called out in the current policy as being allowed.
Jul 1, 2026, 12:08 PM@alexia One of the co-authors has addressed this. I'll try to paraphrase the key points because I think they get lost in the debate:
They (the co-author) have "moral objections to AI" and also "people don't automatically agree what using ai is"
A simple "No AI" rule results in purity tests over tools. Which could grind the project down further, give more traction to pro-AI forks, etc.. So focus on the code, the communication & the contributors.
In reference: https://mastodon.tmm.cx/@hp/116840354608046459
@alexia Possibly more to help reduce disputes over if code is AI that might waste more time than actually reviewing the code.
Not a great situation but they did say their already pretty overwhelmed before AI with submissions as is
Jul 1, 2026, 2:06 PM@alexia I've said it a few times now, but I think the issue is a question of "how substantial is 'substantial'" and the fact that it *needs* to be explicitly stated.
I think the reason for the wording is to account for "AI powered" auto-complete that will guess variable names and common expressions (eg: for loop initialization/conditions). Explicitly rejecting *anything* AI generates kind of locks you into intentionally doing things worse because you saw a better option from an AI by accident.
@bryce@mastodon.brycedixon.dev
I believe this is defined in the full policy document, which namely is not the linked blogpost.
Jul 1, 2026, 2:30 PM@alexia Oh, do you have a link to that document? Its strange they didn't link to it in the post...
@bryce@mastodon.brycedixon.dev
I've not yet taken the time to seek it out, it's probably in a PR somewhere…