GitHub To Use Copilot Interactions To Train AI
From April 24 onward, interaction data from Copilot Free, Pro, and Pro+ users will be used to train and improve our AI models
GitHub has just announced an update to its Copilot usage policy, unveiling plans to utilize Copilot interaction data to train AI.
As of April 24th, GitHub will utilize interaction data from Copilot Free, Pro and Pro+ users to train and improve its AI models. Good news is, they have made it fairly easy to opt out! (Unlike other platforms…)
All you have to do is dive into your Settings > Copilot > Privacy and select “disable.”
They go on to explain that their initial AI models were trained using a mix of publicly available data and hand-crafted code samples but in the past year have been testing with Microsoft employee data, which has shown “meaningful improvements.” Access to this additional data even lead to an improvement in acceptance rates in multiple languages.
In this new training phase, they plan to use:
Outputs accepted or modified by users
Inputs sent to GitHub Copilot, including code snippets shown to the model
Code context surrounding cursor positions
Comments and documentation written by users
File names, repository structure & navigation patterns
Interactions with Copilot features (chat, inline suggestions, etc.)
User feedback on suggestions (thumbs up/down ratings)
Issues, discussions or private repositories at rest will not be used for training, however it will process code from private repos if the user actively uses Copilot.
Business and Enterprise users will not be affected by this update.