Codeberg takes a hard line on AI use with community support

Codeberg’s annual membership meeting tackled two important votes this week, both related to AI.

One formalizes a commitment already implicit in the nonprofit’s privacy policy to not use code or user data to train AI models. The other would rewrite the terms of service to completely ban most AI-generated projects from the platform.

Both proposals passed after a 14-day voting period that ended on July 22nd.

2 votes are

The initial vote turned the existing privacy stance into official policy. Codeberg’s privacy policy already states: Does not require user dataAnd members are now extending the same principles to AI.

Forge and related services We don’t use code or data from your projects or users to train our AI toolsAccording to the motion passed.

The second vote was more hotly contested.. Approximately half of Cordburg’s regular members voted, and the bill passed with a vote of 358 to 144, with 14 abstentions. As a result, new provisions have been added to the Terms of Use (Checkpoint 2.7) Except for projects built primarily from AI-generated code, especially naming tools like Claude and OpenAI Codex.

Why do we do this?

Cordberg’s reasoning goes far beyond a single complaint. The AI ​​crawler accesses the platform indiscriminately and reads every issue filter variant and every point in your Git history. As a result, volunteer system administrators may be required to work on new improvements, resulting in excessive workload.

Hardware costs are also rising. According to Cordberg, drives that cost around 700 euros a few years ago now cost nearly 3,700 euros and are often out of stock. Because it is an association codeberg eVowns the hardware rather than renting cloud capacity, so it has to absorb these price increases directly.

Cordberg also argues that: Vibe-coded disposable software is flooding the platform Projects consume far more storage and CI/CD resources than the number of contributors warrants.

At the same time, maintainers are faced with a growing pile of contributions to review that are generated by low-effort AI.

What does that mean to you?

Cordberg believes: These changes do not result in bulk deletion of existing repositories. Our management team also doesn’t create a purge list, as enforcement is done on a case-by-case basis when issues arise.

Projects with active communities, actual history that predates the AI ​​tool, or the odd AI contribution that is subconsciously accepted are less likely to be affected. Small side projects and experiments are given some leeway as long as they don’t attract too much attention.

The line is drawn for projects built by autonomous AI agentsthe code is primarily maintained by heavy use of AI or by projects that consume far more resources than the number of contributors warrants.

I can’t help but compare this to Linus Torvalds’ recent statements. He stopped calling AI “.90% marketing” Tell the critics to fork the kernel or leave if they don’t like Linux.

What Cordberg has done here is the opposite. The company looked at the same technology and decided that the costs of its servers, trust among contributors, and the model of running a small nonprofit forge outweighed what AI might offer in return.

my humble opinion

I think they did the right thing. Codeberg is only growing in popularity, has active development, and has no shortage of new repositories being built on top of it, so we can’t risk cost overruns at this stage.

Projects like Linux are a different story. It’s one of the most heavily reviewed codebases on the planet. Inappropriate patches generated by AI tend to be discovered quicklyfollowed quickly by some harsh words.