The AI Trap: Why GitHub’s Changes Spell Trouble for Developers

The recent decision by the Dutch Ministry of the Interior to migrate its source code from GitHub to Forgejo, a self-hosted instance, has sparked a wider conversation about the implications of relying on cloud-based platforms like GitHub. As someone who has also made the switch, I want to explore what this move really means for developers and why it’s not just about outages.

The Dutch government’s choice of Forgejo was driven by a desire for digital autonomy, specifically the need to publish its source code on a platform that it owns. This is a crucial distinction: when you host your code on GitHub, you’re essentially giving Microsoft (or any other corporate owner) access to your work. The company can collect and use your data, including interaction data from AI-powered features like Copilot, for their own purposes.

The same reasoning applies to my decision to move away from GitHub. Outages are undoubtedly frustrating, but they’re just a symptom of the larger issue: GitHub’s increasing reliance on AI-driven features. This is particularly evident in its absorption into Microsoft’s CoreAI division, which has further eroded trust among developers. The argument that Microsoft kept GitHub at arm’s length in the past no longer holds true.

When you push code to github.com today, you’re effectively pushing it to a unit of Microsoft’s AI organization. This changes the calculus for many developers who value their independence and want to maintain control over their work. For example, CTO Vlad Fedorov acknowledged that GitHub needs to scale its capacity 30x to keep up with the load generated by new features like Copilot.

The recent announcement about Copilot’s opt-in default is another significant concern. By making interaction data available for model training unless users explicitly opt out, GitHub has flipped the script on user consent. As a maintainer, I shouldn’t have to worry about whether my repository is contributing to AI training; it should be up to me to decide how my code is used.

The implications of these changes are far-reaching. Developers need to think critically about where their code lives and who has access to it. The Dutch government’s decision highlights the importance of digital autonomy and self-hosting. By taking control of our own platforms, we can ensure that our work is not exploited for someone else’s gain.

The shift towards more decentralized and self-hosted solutions could lead to a healthier ecosystem where developers have greater control over their work and its use. This change may also lead to fragmentation of services and a loss of cohesion within the development ecosystem. However, as we look ahead, it’s essential to consider the long-term consequences of relying on cloud-based platforms like GitHub.

The AI trap is real: once you give away control and ownership, it’s difficult to reclaim them. Developers need to be mindful of these risks and take steps to protect their work.