Overview
- GitHub, which announced the change Wednesday, will start using Copilot interaction data from Free, Pro and Pro+ accounts for model training on April 24 unless users opt out.
- The company lists inputs, outputs, code snippets, cursor context, comments and docs, file names and repo structure, feature use such as chat, and thumbs-up or down feedback as data it may collect.
- GitHub says it will not train on private repositories at rest, though code processed while a user is actively using Copilot in a private repo can be logged for training unless that user disables the setting.
- Copilot Business and Enterprise, enterprise-owned repositories, students and teachers are excluded, and prior opt-outs carry over, with a toggle under Settings → Copilot → Privacy for anyone who wants to decline.
- GitHub may share training-use data with affiliates including Microsoft and may retain interaction records for up to five years, drawing sharp pushback from developers and coverage that ranges from critical pieces to step-by-step lock‑down guides.