Jul 29, 2026 07:00:00

Codeberg, a hosting service for open-source software, has banned a generative AI-driven project through a vote.

Codeberg is a non-profit hosting service created for open-source projects and has gained popularity as an alternative to GitHub . Codeberg has decided, through a vote by its members, to ban the 'Vibecode Project,' which is primarily composed of generated AI code.

Protecting our FLOSS commons from LLMs — Codeberg News
https://blog.codeberg.org/protecting-our-floss-commons-from-llms.html

Codeberg Has Drawn a Hard Line on Use of AI With Community Backing
https://itsfoss.com/news/codeberg-bans-ai-contributions/

Codeberg conducted a 14-day vote among its members on two controversial proposals concerning large-scale language models (LLMs). As a result, both proposals were approved by a majority vote, and Codeberg decided to take a tough stance on AI.

The first proposal that was passed was that 'Codeberg will not use project or user data to train AI models.' Codeberg's privacy policy already implicitly indicated that it would not use user data to train AI models. This vote made it clear that Codeberg-related services will not use user data to train AI models.

In a statement, Codeberg said, 'Codeberg forge and related services do not, and will not, use project or user code or data to train AI tools such as LLM (tools that aim to generate output based on training input). The association believes that these technologies are incompatible with responsible creation and maintenance of free and open-source software.'

The second proposal is to revise Codeberg's terms of service to ban 'Vibecode projects' that are primarily powered by generative AI. Here, Vibecode projects refer to those that mainly consist of code output by generative AI such as Claude or OpenAI Codex.

Codeberg cites several reasons for banning the Vibecode project. First, Codeberg points out that LLM is an extremely expensive technology, and as companies enter the investment recoupment phase, the burden on consumers is increasing. This burden will be paid not only by users who actually contract with AI, but also by everyone in the world in the form of rising hardware prices, increased energy consumption, and environmental damage.

For example, Codeberg is regularly subjected to heavy loads from crawlers of AI development companies that try to use the code it hosts for AI training. These crawlers try to read every page hosted by Codeberg, generating costly database queries, which can lead to a decline in service quality and the consumption of time and money spent on building security measures.

Furthermore, Codeberg criticizes developers who use AI to generate massive amounts of code without a development team, creating self-serving projects without a user community. Codeberg stated, 'It seems absurd to us that projects with only one developer and few users are consuming the same or even more resources than the largest community projects running on Codeberg, saving on

CI/CD and storage resources. We don't think it's reasonable to invest our valuable donations in hosting large ghost projects.'

Codeberg also points out that other problems include rising hardware costs due to the increase in AI crawlers and AI developers, the massive energy and water consumption by data centers, and the resulting air pollution and noise pollution to nearby residents.

Codeberg points out that the proliferation of LLMs in FOSS ( Free Software and Open Source Projects) is also an attack on the concepts of trust and friendly collaboration among repository contributors. In fact, the large volume of contributions generated by LLMs submitted to open source projects increases the time and effort that maintainers have to spend on reviews.

The approval of this proposal demonstrates Codeberg's commitment to supporting projects built and maintained by human developers. However, Codeberg understands that many developers use LLMs for their work, and therefore, a large number of repositories will not be deleted immediately.

While it is difficult to give a definitive answer regarding the impact on future projects, it is believed that projects with 'active communities that are interested in and maintain the software,' 'projects with a long history prior to LLM,' and 'projects that do not heavily utilize LLM' are unlikely to be affected.

Furthermore, Codeberg does not plan to spend a great deal of time and resources on automatically scanning all content, so even if it uses AI, it is likely to overlook 'side projects and experiments that consume fewer resources' and 'small tools and custom scripts.'

However, Codeberg stated that projects such as 'projects created autonomously by AI agents,' 'projects created and maintained using LLM extensively,' 'projects that consume significantly more resources than stakeholders,' 'projects deeply involved in the LLM ecosystem (such as tools by LLM developers to facilitate LLM use),' and 'users submitting LLM contributions in violation of the project's custom policies' are likely to be prohibited.