Protecting our FLOSS commons from LLMs
4 hours ago
- Two motions passed: no training LLMs on user data, and terms of use change to prohibit 'vibe-coded' projects.
- Codeberg commits to not using any user or project data for training generative AI.
- LLMs are seen as endangering the free/libre software ecosystem due to high costs, resource strain, and environmental harm.
- Weak-coded projects with single developers and no community consume disproportionate resources, violating Codeberg's mission.
- Rising hardware costs from LLM demand affect Codeberg's ability to expand infrastructure.
- The digital divide widens as costs for digital tools and services increase for small operators.
- LLMs harm collaboration in FLOSS by lowering trust, increasing maintenance burden, and enabling license laundering.
- Enforcement of new terms will be gradual and human-centered, with guidelines for affected projects.