GitHub, the world's largest code hosting platform, recently dropped a major bombshell: it officially announced that starting April 24, 2026, it will use user interaction data to train its AI models. This move has been humorously dubbed a "CTRL-Z" operation by many developers, as GitHub had repeatedly emphasized its commitment to respecting user private data in public statements, and this new policy clearly breaks that understanding.

Code Internet (2)

Controversy over "Default Join", Private Repositories Are No Longer Absolutely Private

According to GitHub's updated privacy policy, users of Copilot's free version, personal version (Pro), and professional enhanced version (Pro+) are all included in this data collection. The system will automatically collect detailed data including code snippets, input and output content, cursor context, and even file names and directory structures. What unsettles the community the most is that even code stored in "private repositories" may be captured for model training if the user enables Copilot while editing. GitHub's Chief Product Officer Mario Rodriguez explained that internal employee tests have shown that adding real interaction data significantly improves the AI's accuracy in detecting bugs. However, this "default on" rather than "manual opt-in" strategy has triggered strong backlash in the developer community, with opposition votes skyrocketing almost instantly under the announcement.

How to Protect Yourself: Exemptions for Enterprise Users and Manual Deactivation Guide

In this data harvesting battle, not all users are in a passive position. GitHub explicitly stated that paid organizations who purchased Copilot Business and Enterprise versions, as well as certified students and teachers, will be protected by contract terms, and their data will not be used for training.