GitHub has announced an update to how it handles data generated through its AI coding assistant. From 24 April 2026, all user interactions with GitHub Copilot (including prompts, code snippets, outputs, related context etc) will be used to train and improve its AI models unless users choose to opt out.
Users are being encouraged to review their settings now this change has taken effect. To opt out:
- Go to GitHub Account Settings
- Select Copilot
- Choose whether to allow your data to be used for AI model training
Those who previously disabled data collection for product improvement purposes will retain that preference, so their data will not be used for AI training unless they actively opt in.
GitHub states that the update is intended to improve the quality and relevance of Copilot’s suggestions by making them more context-aware, and points to internal testing suggesting improved acceptance rates across multiple programming languages. The change reflects a broader industry trend towards using real-world user interactions to refine AI systems. In principle, this should lead to make accurate and useful assistance.
However, it also raises ongoing questions around consent and transparency for developers. Copilot interactions may include proprietary or sensitive code and it is not always clear how such material may be processed, stored, or separated from training datasets. Although opt-out controls are available, some users may question whether the default position sufficiently protects those working of private or commercially sensitive projects, particularly where data protection obligations apply.