Codeberg bans “vibe-coded” AI projects, citing FLOSS harm and unclear copyright
Berlin’s volunteer-run host votes to block most generative-AI-written code and refuses AI training use of users’ data.

Codeberg e.V., the Berlin-based non-profit behind Codeberg, has voted to ban “vibe-coded projects” and will not use users’ code or data for AI training due to alleged Free, Libre, and Open Source Software (FLOSS) impacts. The move includes amended Terms of Use and follows internal concerns about bot traffic, rising infrastructure costs, and trust in community collaboration.
Codeberg just drew a line in the sand for AI-assisted software, and it is not a vibes-only debate. On Thursday, Codeberg e.V., the Berlin-based non-profit that oversees the code hosting service, voted to ban “vibe-coded projects” and said Codeberg will not use users’ code or data for AI training because of what it believes is the damage to Free, Libre and Open Source Software (FLOSS).
For anyone building, hosting, or funding open source, the immediate consequence is blunt: projects “mostly consist[ing] of code written by 'generative AI'-tools (including services such as Claude, OpenAI Codex)” are no longer welcome at Codeberg. The Terms of Use also point to a core risk the community is worried about: “unclear copyright status” plus “little safeguards to ensure that they do not include harmful code.”
So what triggered this vote? According to Codeberg members, it is not just philosophical distaste for AI coding. It also starts with cost and capacity, and then morphs into trust and governance. The Register reports that Codeberg has been trying to prevent automated software (bots) from taxing its infrastructure with excessive network requests. In parallel, the organization now says AI companies shift costs to others, and those costs show up in very real operational terms for a small, volunteer-run platform.
In a blog post, authors Bastian Greshake Tzovaras, Otto Richter, and William Zijl argue that “LLMs are so costly that companies externalize the costs on a massive scale - on those who don't use them and society at large.” They connect that to “increased hardware prices, energy use and environmental damage - we all pay for it!” They even cite Codeberg’s SSD and memory hardware: a drive that only a few years ago cost €700 (~$800) now costs €3,700 (~$4,200), if it is even in stock. The result is that Codeberg has been forced to raise prices.
But the “infrastructure bill” framing also expands into a community dynamic. The authors call out a wave of projects that involve a solo developer “working with a statistical machine that turns energy into code.” The criticism here is not that the output exists. It is that it does not participate in the social contract that FLOSS depends on: community, maintainers, and ongoing stewardship. Codeberg argues that its limited CI/CD and storage resources get eaten by “ghost projects” that never build a real user or contributor base.
Then there is the trust angle, which is where this stops being a cost debate and becomes a governance debate. The piece reports that Codeberg members believe LLM use is undermining what makes FLOSS work: “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,” they state. The stated mechanisms are practical and operational: LLMs “magnify maintainers' workloads,” create confusion about whether projects will be maintained, and contribute to what the authors call “license laundering,” where copyleft code is stripped of its reciprocity requirements by “generating” it out of training data.
What makes this decision notable is how it is being enforced on paper even if enforcement in practice may be limited. Codeberg’s statement about limited resources and an overburdened workforce implies that policing every repository may not be feasible unless a project “draws attention to itself.” Still, the Terms of Use change provides a legal and community governance hook: the rule is explicit about what users must not share and why. That means other hosting providers, foundations, and even enterprise customers who depend on open source provenance now have a new precedent to consider when they decide where to host AI-assisted code and how to assess licensing and security.
The vote itself suggests internal support, but not unanimity. The Register reports 358 voted in favor, 144 voted against, and 14 abstained, with about half of the active members voting. The backlash is also documented: Armin Ronacher, creator of Flask and co-founder of AI agent biz Earendil, is quoted saying, “I think this is a very bad move, and the people behind Codeberg should re-consider their stance.” Whether you see the ban as protection or gatekeeping, it is clearly a meaningful moment for the open source “commons” at a time when more tooling is being generated automatically and distributed faster than maintainers can keep up.
Second-order implications are going to show up in boardrooms, not just repos. Codeberg also decided to ban cryptocurrency projects, citing a similar move by SourceHut in 2023. On the “purpose” front, a proposed amendment to make it Codeberg’s stated purpose “to oppose discrimination and promote a diverse FOSS community” passed with a two-thirds vote but is not yet merged. In other words, this is a platform defining which values it can operationalize. For founders and investors watching developer infrastructure, the stake is clear: as AI code generation accelerates, community-hosting policies become a competitive differentiator, and they can reshape where innovation lands, where costs accrue, and who gets to participate in the FLOSS supply chain.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Technology

Microsoft and Windows users caught LG installing McAfee ads via Windows Update
LG monitor software piggybacks on Windows Update, then surfaces McAfee free-trial pop-ups, prompting a quick public backlash and response.

AMD’s Helios racks 72 MI455X accelerators as it targets Nvidia lead in AI
At its San Francisco Advancing AI event, AMD pitches MI455X and Helios to win both training and inference performance.

Musk wants frontier AI peer-reviewed by rivals, not regulators, and says do it immediately
Fortune reports Musk’s plan: weekly rival calls for safety checks, with government only stepping in when a firm can’t fix issues.

