The Godot Foundation recently announced stricter rules for accepting changes to the open game engine Godot. The foundation has decided to completely ban the use of autonomous AI agents or vibe coding, as well as the generation of significant code fragments using AI. It is now mandatory that all code be created by a human, with AI only permitted for routine development operations such as code completion and find/replace. Any use of AI in writing code must be disclosed in the pull request discussion.
In addition, AI is not to be used for reviewing pull requests – all submissions must be reviewed and approved by humans before being accepted. The use of AI to generate text for explanations, proposals, problem reports, and other forms of human-to-human communication is also prohibited. However, AI can be used for machine translation into another language as long as a human writes the translated text.
To prevent the hidden use of AI and reduce the burden on maintainers, the project has established restrictions for new developers. Those with fewer than four accepted pull requests will not be allowed to propose new functionality or carry out major code refactorings without prior approval from maintainers. This measure aims to encourage newcomers to familiarize themselves with the codebase, fix bugs, and work on documentation to earn the trust of maintainers before embarking on significant projects.
The decision to ban AI in Godot development was primarily motivated by maintainers becoming overwhelmed with the increasing number of questionable changes produced with AI assistance. Currently, the Godot repository on GitHub has accumulated more than five thousand open pull requests that have yet to be addressed. The rise of AI tools has led to a surge in pull requests, despite the limited number of reviewers available. Reviewers are now spending valuable time evaluating changes that often lack coherence, have verbose descriptions, and are submitted by developers who may not fully comprehend or test their own patches.