Codeberg has made its choice. The Berlin-based nonprofit that runs a popular open-source code hosting service just voted to push out projects built mostly by large language models. No more vibe-coded repos cluttering its servers. No more letting AI drain resources built for people.
The Vote That Changed the Rules
Members of Codeberg e.V. cast ballots this month. The tally showed 358 in favor, 144 against, and 14 abstentions. Roughly half the active members weighed in. The outcome produced two clear policies. First, Codeberg promises it will never feed user code or project data into LLMs for training. Second, it amended its terms of use to ban sharing projects that mostly consist of code written by generative AI tools.
The new language lands with precision. “You must not share projects that mostly consist of code written by ‘generative AI’-tools (including services such as Claude, OpenAI Codex). Such projects having an unclear copyright status … and furthermore have little safeguards to ensure that they do not include harmful code.” Those words come straight from the commit that updated the Terms of Use.
But why now? The reasons stretch beyond one vote. For years Codeberg fought an uphill battle against automated bots hammering its infrastructure. Last summer it tightened defenses after AI agents bypassed existing protections. The pattern became impossible to ignore. Solo developers, often working alone with an LLM, upload vast amounts of code that sees little human review. These ghost projects consume CI/CD cycles, storage, and volunteer time. All while delivering questionable value to the broader FLOSS community.
Costs tell part of the story. Hardware prices have climbed sharply. A solid-state drive that once ran €700 now sits near €3,700 when available at all. Energy demands rise in parallel. Water usage for data centers adds another burden. Codeberg doesn’t operate in a vacuum. It sees these pressures firsthand and refuses to subsidize them. “LLMs are so costly that companies externalize the costs on a massive scale – on those who don’t use them and society at large,” wrote the authors of the organization’s explanatory blog post. Increased hardware prices, energy use and environmental damage – we all pay for it!
The critique runs deeper than dollars. Trust sits at the center of open source. Maintainers rely on clear signals about who wrote what and whether a project will receive ongoing care. AI output clouds that picture. It multiplies review workloads. It raises doubts about long-term maintenance. And it opens doors to license laundering, where copyleft protections get stripped away because the model regurgitates training data without proper attribution.
“The widespread use of LLMs in FLOSS is instead becoming a multidimensional attack on the trust between contributors and the very idea of convivial collaboration itself,” the post declares. Strong language. Yet it matches the frustration many maintainers voice in private forums and public threads.
Enforcement won’t prove easy. Codeberg admits its small team already stretches thin. The policy likely triggers only when a project draws attention through complaints or obvious patterns. Still, the signal matters. Projects built primarily by AI now know they should look elsewhere. The organization points them toward other hosts without apology.
Reactions split along predictable lines. Armin Ronacher, creator of the Flask web framework and co-founder of an AI agent company, didn’t hold back. “I think this is a very bad move, and the people behind Codeberg should re-consider their stance,” he posted on X. Others praised the decision as a necessary defense of human-centered development. Discussions on Hacker News, Reddit’s selfhosted forum, and various developer Discords lit up within hours of the announcement.
Codeberg didn’t stop at AI. It also voted to ban cryptocurrency-related projects, citing similar resource and philosophical concerns that drove SourceHut’s 2023 decision. A separate proposal to explicitly oppose discrimination and promote diversity in its community passed with a two-thirds majority but awaits final merging. The organization continues refining its identity as a thoughtful alternative to corporate forges.
This move lands at a tense moment for open source. Major platforms have embraced AI assistance. GitHub Copilot and similar tools now sit inside countless workflows. Yet questions about copyright, provenance, and sustainability refuse to fade. European jurisdictions add complexity. Codeberg operates under German law, where the copyright status of purely machine-generated work remains unsettled. The organization prefers caution over risk. Better to require meaningful human contribution than host code whose legal standing could unravel later.
Recent coverage echoes these tensions. Codeberg’s own blog post lays out the full reasoning in careful detail. LWN.net summarized the policies and noted the controversy around the hosting ban. XDA Developers framed the decision as Codeberg picking a side in the widening open-source AI debate. And LinuxStans questioned enforcement practicality while acknowledging the copyright motivations.
Discussions on X captured the divide. Some users called the policy ideological overreach. Others saw it as a principled stand against cost externalization. One developer noted the rule effectively covers bases under German copyright law, ensuring hosted material carries clear human authorship and license compliance. Another highlighted that small amounts of AI assistance remain acceptable. The ban targets projects where AI does the heavy lifting with minimal human oversight.
So what does this mean for developers? Those who treat LLMs as a thoughtful assistant rather than a replacement face little disruption. Review your code. Understand it. Maintain it. Document decisions. These steps satisfy both the letter and spirit of the policy. But teams that generate entire applications through prompts and upload without scrutiny now risk account action.
The decision also spotlights larger supply-chain questions. Enterprises that consume open-source dependencies must track provenance more carefully. If a popular library incorporates unreviewed AI output hosted elsewhere, downstream risks multiply. Security vulnerabilities become harder to trace. License obligations grow murky. Codeberg refuses to become part of that problem.
Its stance aligns with a growing cohort of independent forges and maintainers who prioritize sustainability over volume. SourceHut’s earlier crypto ban set one precedent. Smaller self-hosted instances multiply as developers seek control. The fragmentation carries trade-offs. Yet many view it as preferable to depending on platforms that monetize developer data or push AI features without community consent.
Codeberg built its reputation on independence. Funded by donations and run by volunteers, it answers to its members rather than shareholders. That structure allowed this vote to happen. It also limits the organization’s scale. The same resource constraints that prompted the policy now shape its enforcement. Expect pragmatic application. Clear human stewardship will likely outweigh technical violations in gray areas.
Still, the message resonates. Open source thrives on human relationships, shared understanding, and mutual accountability. When statistical models flood the commons with low-effort output, those qualities erode. Codeberg has decided not to accelerate that erosion. Human collaboration first. Simple. Direct. And, for this community, non-negotiable.
The coming months will test whether other hosts follow suit or double down on AI integration. For now, Codeberg stands apart. Its members spoke. The rules changed. Developers who value the same principles now know where they can build without subsidizing the alternative.
WebProNews is an iEntry Publication