Codeberg, a non-profit code-sharing platform hosting many popular free and open source projects, is banning AI-generated, a.k.a. “vibecoded,” projects. The community also pledges not to use any user data for AI training.

Codeberg’s updated policy explicitly prohibits AI-generated projects.

“You must not share projects that mostly consist of code written by ’generative AI’-tools (including services such as Claude, OpenAI Codex),” the updated terms of service reads.

Failure to comply results in the immediate content removal, and subsequent violations might result in account suspension.

While the policy briefly cites “unclear copyright status” and potentially “harmful code” as the reasons, Codeberg dropped a sharply worded statement that explains the reasoning in more detail. Codeberg blames AI tools for increasing platform load and costs, eroding collaboration and trust, and damaging the broader environment.

The sweeping majority of Codeberg’s active members supported the decision to ban AI projects: 358 agreements, 144 disagreements, and 14 abstentions.

“To us, it seems ridiculous to see projects with a single developer and virtually no users consuming as much or even more resources than some of the largest community projects on Codeberg,” the organization explained in a blog post.

“We do not believe it is reasonable for Codeberg to invest our precious donation money into hosting of large ghost projects.”

Codeberg says it doesn’t want to be a dumping ground for generated, single-use software.

Some forms of AI slop will be “tolerated”

Codeberg, however, doesn’t plan to spend any significant resources to automatically scan content for AI use. This means that small projects and experiments with little resource use will be tolerated in practice.

The team explains that the platform welcomes projects with an active community maintaining free software, especially with a significant pre-LLM history. Maintainers, who accidentally or willingly accept LLM-generated contributions, won’t be penalized, as long as the development does not involve “heavy use of LLMs.”

“You won't see a mass-deletion of content within days,” Codeberg assured.

“We’re humans at the other end, who care deeply about free/libre software projects and communities.”

The projects that no longer have a future on the platform are as follows:

• Autonomously created by AI agents.
• Written and maintained with heavy use of LLMs.
• The amount of resource use (storage, CI/CD) is significantly larger compared to what the involved maintainers could have created by hand.
• Projects heavily tied to the AI ecosystem, such as tools that make using LLMs easier.

Users who send LLM contributions in violation of the project’s policies are also no longer welcome.

Codeberg compares vibecoders to “the development team of none,” spending tokens for a kick of adrenaline, using a statistical machine to rapidly turn energy into code. Some projects have heavy code activity, extensive CI/CD testing, frequent and large binary releases, and more supported platforms than actual users.

“It seems like many ‘vibecoders’ don’t realize that they don't actually have a community around them. They build projects as if they had, and spend resources accordingly.”

AI is turning personal computing into a luxury

A lot of the blog post is dedicated to AI-induced direct and indirect costs.

“LLMs are a very costly technology, and those costs keep rising as the companies providing them have to start recouping their investments,” the organization said.

Web crawlers are already straining servers as they try to read every page, resulting in expensive database queries and diminished service quality for legitimate users.

Codeberg also noted that the same storage device (SSD), which used to cost €700 a few years ago, now costs €3,700, and is often out of stock. The expansion of data centers and AI is driving up RAM and storage prices, which in turn is making hosting more expensive, leading to stricter quotas for users.

The organization also warns of a growing digital divide, as only the largest tech companies have reliable hardware agreements.

“Even more basic digital tools like computers and smartphones are affected by the rising costs, turning personal computing back into a luxury,” Codeberg said.

Due to AI expansion, customers can’t afford decent devices and are being pushed into the traps of cloud providers, while also paying higher electricity and water prices.

Codeberg also sees AI eroding collaboration and trust – people use AI to write single-use throwaway software from scratch. Generated code also strips copyleft licensing from code it was trained on, leading to “license laundering.”

While some people try to present slop as a real effort, those who put in actual work are often falsely accused of using AI.

Maintainers are under an increased workload due to people submitting (often well-meaning) low-effort, LLM-generated contributions that require substantial amounts of time to review

Codeberg

The organization fears that the vicious cycle leads to less rewarding collaboration, in which people are less likely to write high-quality software.

“Sharing the result of a prompt … does not make the world a better place.”