GitHub to Train AI Models on User Data Starting April 24

GitHub has announced a significant policy change that will allow the company to use customer interaction data to train its AI models starting April 24, 2025. The revision affects Copilot Free, Pro, and Pro+ customers, while Copilot Business, Copilot Enterprise, students, and teachers remain exempt under existing terms.

The change means that unless users actively opt out, their inputs, outputs, code snippets, and associated context will be collected and used to improve GitHub's AI capabilities. To opt out, users must visit /settings/copilot/features and disable the "Allow GitHub to use my data for AI model training" option under Privacy settings.

What Data Will Be Collected

According to GitHub's documentation, the collected data includes:

• Model outputs that have been accepted or modified
• Model inputs including code snippets shown
• Code context surrounding your cursor position
• Comments and documentation you've written
• File names and repository structure
• Interactions with Copilot features (e.g., chats)
• Feedback such as thumbs up/down ratings

Industry Context and Comparisons

GitHub defends the policy shift by noting that Anthropic, JetBrains, and Microsoft operate similar opt-out data use policies. The company argues that interaction data significantly improves AI model performance, citing increased acceptance rates for AI model suggestions after incorporating Microsoft employee interaction data.

However, the policy raises questions about the true meaning of "private" repositories on GitHub. While private repositories are described as "only accessible to you, people you explicitly share access with, and, for organization repositories, certain organization members," the new policy effectively creates a asterisk on that definition. Code snippets from private repositories can be collected and used for model training while users are actively engaged with Copilot in those repositories.

Community Response

Initial reactions from the GitHub community have been largely negative. Emoji voting on the announcement shows 59 thumbs-down votes compared to just three rocket ship emojis indicating excitement. Among 39 comments at the time of publication, only GitHub's VP of Developer Relations Martin Woodward has publicly endorsed the change.

Broader Implications

The policy shift highlights a fundamental tension in the AI industry: the reliance on vast amounts of data for model training versus user privacy expectations. GitHub's FAQ notes that OpenAI's Codex, which powers GitHub Copilot, was "fine-tuned on publicly available code from GitHub," suggesting that data collection for AI training has been occurring in various forms for years.

For developers and organizations concerned about data privacy, the April 24 deadline provides a clear timeline for action. Those who wish to maintain control over their code and interaction data should review their GitHub settings before the policy takes effect.

Compliance and Legal Considerations

The opt-out approach aligns with US industry norms rather than European standards, which typically require opt-in consent for data processing. This distinction is particularly relevant for international organizations and developers working across jurisdictions.

The policy change also raises questions about liability and intellectual property rights, especially as AI models become increasingly sophisticated at generating code that may resemble or replicate existing patterns from training data.

For more information about GitHub's data usage policies and how to opt out, visit GitHub's official documentation.