When people think about private data, they think about names, emails, photos, health records. Source code rarely makes the list. It should.
A code repository is not only the app. It holds the whole history of how the app was built: every change, every comment, every message a developer wrote at two in the morning. It shows how the product works inside, where the weak spots were, which customers asked for which feature, and very often a few things that should never have been committed at all. If you pay someone to build software for you, that repository is part of your company.
So the question “where does the code live?” is a privacy question. Here is how we answer it, and why.
Big hosts belong to big companies
Most code in the world sits on a few large platforms. GitHub has belonged to Microsoft since 2018, when Microsoft bought it for $7.5 billion. Bitbucket has belonged to Atlassian since 2010. These are good products built by good engineers. They are also parts of large companies with their own business plans, and those plans change.
Here is a recent example. In March 2026 GitHub announced that from April 24 it would use interaction data from its AI coding assistant to train its models, unless the user opts out. That data includes code snippets sent to the assistant, the code around the cursor, file names and the structure of the repository. It may be shared with Microsoft. Business and enterprise plans are excluded, and GitHub says it does not train on private repositories “at rest”. But code from a private repository that a developer works on with the assistant switched on is fair game, unless someone found the setting and turned it off.
None of this is secret or illegal. It was announced in a blog post, and the default was set to “yes”. That is the point. The rules for your code are written by someone else, and they can be rewritten with a blog post. You find out after the fact, if you find out at all.
We can’t audit what happens to code on someone else’s servers. Nobody outside those companies can. So we don’t put our code there.
What we do instead
We run our own Git server. Our code, its history and our build pipelines live on machines we control, in the EU, under our own rules. Apps are built by our own build machines, including a Mac for the Apple builds. No AI training setting to find, no terms of service to re-read every year, and no surprise when a platform decides to “improve its models”.
This is the same idea we apply to user data in our apps: data that is never handed over cannot be misused. If we would not upload your notes to a third party, it would be odd to upload the code that handles them.
It is not free. Someone has to keep the server updated, watch the backups and fix things when they break. For us that cost is small and it buys a clear answer to “who can see this?”: us, and nobody else.
GitHub is not the obvious home for open source anymore
Open source is different. The code is public by design, so privacy is not the issue. The issue is whether the platform still serves the project.
More and more maintainers say it doesn’t. In November 2025 the Zig programming language moved its main repository from GitHub to Codeberg. Its lead developer listed the reasons in public: a slow and buggy web interface, a build system with serious bugs that was left unfixed, and GitHub pushing AI features at a project that does not accept AI-written contributions.
Codeberg is run by Codeberg e.V., a non-profit association based in Berlin. Its members decide the budget, and its mission is free and open-source software, not selling something else to the people who use it. For an open project that is a better fit: no ads, no upsell, no corporate parent with other goals.
One honest note: Codeberg is for open source. For private company code its own documentation suggests running the same kind of software on your own server. Which is exactly what we do.
Where GitHub still wins
It would be unfair to stop here. GitHub has real advantages, and one of them is big for app makers: build machines.
GitHub runs builds for public repositories for free on its standard machines, and that includes macOS. Any app for iPhone or Mac has to be built on a Mac, and Macs in the cloud are expensive. For private projects GitHub charges about six cents a minute for a standard macOS machine, roughly ten times its Linux price. An app that builds for twenty minutes, several times a day, adds up fast. The other route is buying and running your own Mac, which is what we do.
So for a small open-source app, especially one that ships on Apple platforms, GitHub is still the cheapest way to get it built. Its reach counts too. Most developers already have an account there, which makes it easy to get bug reports and help.
Our own rule is simple:
- Private code (our apps, client work): our own server, no exceptions.
- Open source: public anyway, so the choice is about cost and community, not secrecy. Free Mac builds are a fair reason to use GitHub. Codeberg is the better home when they don’t matter.
What to ask the people who build your software
If a studio or freelancer builds software for you, ask them three things:
- Where does my code live? On whose servers, in which country, under whose terms?
- Can it be used to train AI models? If they use a hosted platform or an AI assistant, which settings are on, and who checked?
- Can I get all of it, history included, and take it elsewhere? You should own the repository, not only the final app.
A good partner answers all three in a minute and without changing the subject. Your code is your data. Treat its home the way you would treat any other place where your company’s secrets are kept.