Codeberg Shuts Out Crypto Projects and AI Written Code in Major Open Source Shift
Open source hosting is changing fast.
Why the New Rules Came In
The site held a vote among its users. The first question was about stopping people from scraping code to train AI models. This passed easily. The second question asked if projects built mostly with AI tools like Claude or OpenAI Codex should be allowed. The answer was no, with 358 votes in favor of the ban and 144 against.
These changes are now part of the site’s terms. They cover both crypto work and AI-generated code. The goal is to keep the platform fair and sustainable for real human developers.
Problems With Cryptocurrency Projects
Many crypto projects take up server space and bandwidth without giving much back to the community. Codeberg saw rising costs from these projects. At the same time, the projects often focus on quick profits rather than long-term open source values. The ban helps the site stay true to its roots as a place for genuine collaboration.
Issues With AI-Generated Code
AI tools create new headaches. They can copy code from many places and hide the original licenses. This is called license whitewashing. It makes it hard to know who really owns the work. On top of that, running AI models needs huge data centers. These centers drive up hardware prices for everyone, including small sites like Codeberg.
Studies also show that heavy use of AI coding tools can hurt the skills of developers. When people stop writing code themselves, the whole open source world loses creative energy. Banning scraped data for training while still allowing AI output would not be fair. So the site chose to stop both.
What This Means for Developers
Real coders who write their own work can keep using the platform without worry. Projects that mix some AI help with human review might still be okay if the human part is the main part. But anything that is mostly machine-made will be removed.
This move could push other hosting sites to think about the same issues. It also reminds everyone that open source works best when people share their own ideas and respect each other’s licenses.
Looking Ahead
The decision shows that community sites can set their own standards. As AI tools grow, more platforms may draw similar lines. For now,