GitHub Will Use Copilot Interactions to Train AI Models Starting April 24

GitHub has quietly updated its Copilot usage policy with a significant change: starting April 24, 2026, all interactions with GitHub Copilot will be used to train Microsoft's AI models.

The change applies to all GitHub Copilot users, with Microsoft justifying it as a way to continuously improve its models based on real-world usage. Users who do not wish to contribute to training can opt out in settings.

The practice isn't new in the industry. OpenAI, Google, and other AI providers have long used user interactions for model training, but the GitHub change is notable because it affects millions of developers using Copilot daily in professional contexts.

For companies with strict data handling policies, this means reviewing their GitHub agreements and considering whether Enterprise licenses with stronger data protections are necessary.

What does this mean for developers?

Code, comments, and prompts you write in Copilot could potentially become training data. Microsoft has not provided details on how data is anonymized or what guarantees apply to confidential code.

It's worth noting that GitHub Enterprise customers have historically had stronger data protection guarantees. Review your current Copilot plan and associated terms before the April deadline.