GitHub has announced that starting April 24, they will begin collecting Copilot interaction data from private repositories to train AI models for users who haven't opted out. The policy affects Free, Pro, and Pro+ accounts, while Business and Pro subscribers remain exempt.

The collected data includes inputs, outputs, code snippets, and associated context from Copilot usage. GitHub clarified they won't train on "private repo data at rest," but the distinction between repository content and interaction data raises concerns about what constitutes truly private code.

Privacy advocates criticize the opt-out approach as unethical, arguing that active consent should be required. The policy change has reignited discussions about data ownership in cloud development platforms, with some users considering self-hosted alternatives to maintain control over their intellectual property.