78% of Open Source Projects Now Allow AI-Generated Contributions, Study Finds
New research on 1,000 GitHub repos reveals how GenAI policies are shaping open source contributions.
A recent paper from arXiv (2605.16706) by Andre Hora and Romain Robbes provides the first empirical study on how open source projects are updating contribution guidelines in response to generative AI. By analyzing 1,000 popular GitHub repositories, the researchers identified 118 repositories with explicit AI policies for contributors. The results reveal a nuanced landscape: 78% of policies permit AI-assisted contributions, while 22% explicitly discourage or prohibit them. This openness, however, comes with strings attached — 51% require contributors to disclose when code was generated or assisted by AI, and a striking 74% demand that a human remains in the loop throughout the contribution process.
The study underscores that the open source community is not simply accepting AI-generated code without oversight. The human-in-the-loop requirement suggests projects value accountability and quality control, likely due to concerns about license compliance, originality, and potential bugs from AI-generated code. The disclosure mandate aligns with broader trends in transparency, helping maintainers review contributions appropriately. For developers and researchers, these findings imply that contributing to open source with AI assistance is increasingly viable, but only if you're transparent about it and ready to demonstrate human oversight. As GenAI continues to evolve, these policies may become a template for collaborative software development.
- 78% of 118 AI policies in top GitHub repos allow GenAI contributions; 22% discourage them
- 51% of policies require explicit disclosure when contributions are AI-generated
- 74% mandate a human in the loop, ensuring human oversight of AI-assisted code
Why It Matters
For open source contributors: AI code is welcome—but only if you disclose it and keep a human in charge.