Starting April 24, 2026, GitHub will use "interactions with GitHub features and services, including inputs, outputs, code snippets, and associated context" from Copilot Free, Pro, and Pro+ users to train AI models. That language comes straight from GitHub's updated documentation. If you don't actively flip a toggle in your settings before that date, your data is in the training pool.
Notice the carve-out: Copilot Business and Enterprise customers are explicitly excluded. GitHub's docs state that "Copilot Business or Copilot Enterprise customer data" is "protected under GitHub's Data Protection Agreement, which prohibits such use without customer authorization." If you're paying $19/month for Pro, your code interactions feed the model. If your company is paying $19/seat for Business, they don't. Same price. Different data rights.
What "interaction data" actually means
GitHub's documentation defines the scope broadly: inputs, outputs, code snippets, and associated context. That covers your prompts to Copilot Chat, the code suggestions Copilot generates for you, the code you were writing when those suggestions appeared, and whatever surrounding context the system used to produce them.
This is not just your public repositories. This is your active development workflow, the questions you ask an AI coding assistant, the code you accept and reject, and the patterns of how you write software. For developers working on proprietary projects using personal Copilot subscriptions, this should raise immediate questions about what ends up in the training set.
GitHub frames the change as building "more intelligent, context-aware coding assistance based on real-world development patterns." That's a reasonable technical goal. But the mechanism for achieving it, opt-out rather than opt-in, is the policy choice worth examining.
The opt-out problem
The opt-out toggle lives in your Copilot settings on GitHub.com. Click your profile picture, go to Copilot settings, and look for the model training section. GitHub says you can "opt-out from allowing your data to be used for training in your personal settings."
The problem is behavioral, not technical. Research on default settings consistently shows that the vast majority of users never change defaults. When the EU's GDPR required opt-in consent for data processing, the number of people "agreeing" to data collection dropped dramatically compared to opt-out regimes. GitHub knows this. Every product team at every major tech company knows this. The default is the decision for most users.
GitHub Copilot Free launched in late 2024 and opened the service to anyone with a GitHub account. That means a massive pool of individual developers, many of them students, hobbyists, and early-career engineers, will have their interaction data swept into training without ever knowing the policy changed.
