GitHub Copilot to Use User Code for Training

GitHub Copilot will now use user code for AI training, with an opt-out option for Free and Pro users starting April 24.

2 min read
GitHub Copilot to Use User Code for Training
Github Blog

GitHub is updating its data usage policy for GitHub Copilot, allowing the AI coding assistant to leverage user interaction data for training its models. Starting April 24, inputs, outputs, code snippets, and associated context from Copilot Free, Pro, and Pro+ users will be collected for model improvement unless users actively opt out. This move aims to enhance the AI's accuracy and contextual awareness, drawing parallels to earlier improvements seen from using Microsoft employee data.

The policy change, detailed in a blog post, specifies that data such as accepted or modified suggestions, code context, comments, and interaction patterns will be used. This aligns with industry trends in leveraging real-world data to refine AI capabilities. Users who previously opted out will retain their preferences, ensuring their data is not used unless they choose to opt back in.

Crucially, this policy update does not impact Copilot Business or Copilot Enterprise users. Their data is not subject to this new training regimen. GitHub emphasizes that the collected data will be shared only with GitHub affiliates, including Microsoft, and not with third-party AI model providers.

Opt-out is Key

While GitHub frames this as a way to build "smarter models" and improve the overall developer experience, the opt-out mechanism is critical. Developers concerned about their code, even snippets, being used for training must navigate to their Copilot settings to disable data collection for product improvements. This data collection is distinct from the operational data required to run the service, which inherently processes code from private repositories.

The company believes that real-world interaction data is essential for the future of AI-assisted development. This GitHub Copilot updates initiative is part of a broader push to integrate AI more deeply into the LLMs developer ecosystem. Similar advancements are seen in how GPT-5.3 Codex powers GitHub Copilot, Cursor, and the ongoing development of AI agents like those in GitHub's Squad: AI Agents Live in Your Repo.

© 2026 StartupHub.ai. All rights reserved. Do not enter, scrape, copy, reproduce, or republish this article in whole or in part. Use as input to AI training, fine-tuning, retrieval-augmented generation, or any machine-learning system is prohibited without written license. Substantially-similar derivative works will be pursued to the fullest extent of applicable copyright, database, and computer-misuse laws. See our terms.