GitHub Coding Agent : Data retention for prompts and outputs in GitHub Coding Agent #183099
Replies: 1 comment
-
|
Based on the official privacy documentation and recent updates regarding GitHub Copilot Business, Enterprise, and Pro plans, here is the clarification regarding data retention for the Coding Agent and "outside the IDE" tools. 1. Does GitHub Coding Agent retain prompts and data for 28 days?Yes, and potentially longer depending on the context.
2. Is Coding Agent treated the same as “outside the IDE (CLI)”?Yes, in terms of not being "Zero Retention." 3. Is it acceptable for Enterprise environments?Yes, generally.
Summary Table for Business/Enterprise Plans
Feature | Data Retention Policy | Used for Training?
-- | -- | --
Copilot Completions (IDE) | Zero Retention (Discarded immediately) | No
Copilot Chat (IDE) | Zero Retention (Discarded immediately)* | No
Copilot CLI | Retained for 28 Days | No
Copilot Coding Agent | Session Logs Retained (Life of account/session) to allow user review | No
*Note: Chat history is stored locally on your machine or in your conversation history UI for your convenience, but is not retained by GitHub's backend processors for training.[5] Recommendation: If your organization requires strict "ephemeral only" interactions where no data ever persists, you should stick to standard Copilot Completions. However, for the Coding Agent to function (planning and executing complex tasks), retaining the "state" of the session is functionally required, and this data is protected under your standard Enterprise agreement.[5] Google Search Suggestions Display of Search Suggestions is required when using Grounding with Google Search. Learn more
Yes, and potentially longer depending on the context. Copilot in the CLI: For Business/Enterprise plans, prompts and outputs generated via the CLI are typically retained for 28 days.[1][2][3][4] This is distinct from the IDE extension to allow for abuse monitoring and service stability in the command-line environment. Copilot Coding Agent (Agent Mode / Workspace): Because the Coding Agent performs multi-step autonomous tasks (planning, editing multiple files, running commands), it generates Session Logs.[5] Retention: These session logs (the plan, the steps taken, and the output) are often retained for the life of the account (or until the user deletes the session). Reason: This is not for GitHub's training purposes, but to provide the service to you. Unlike a fleeting code suggestion, an Agent session is a persistent record (like a Pull Request or an Issue) that you and your team need to be able to review, audit, and refer back to.[5]
Yes, in terms of not being "Zero Retention."
Yes, generally. No Model Training: Crucially, for Business and Enterprise plans, GitHub does not use your prompts, session logs, or outputs to train their public AI models, regardless of whether it is the Coding Agent, CLI, or standard Copilot.[5] Data Isolation: The retention (28 days or session history) is strictly for providing the service (allowing you to view the agent's history/logs) and compliance/abuse monitoring. Enterprise Control: The data remains isolated to your organization. If the Coding Agent creates a Pull Request or an Issue, that data persists just like any other code or comment your developers push to GitHub. Summary Table for Business/Enterprise Plans *Note: Chat history is stored locally on your machine or in your conversation history UI for your convenience, but is not retained by GitHub's backend processors for training.[5] Recommendation: If your organization requires strict "ephemeral only" interactions where no data ever persists, you should stick to standard Copilot Completions. However, for the Coding Agent to function (planning and executing complex tasks), retaining the "state" of the session is functionally required, and this data is protected under your standard Enterprise agreement.[5] Sources |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Select Topic Area
Question
Body
Question: Data retention for prompts and outputs in GitHub Coding Agent (Business/Enterprise/Pro plans)
Hi GitHub team,
I have a question about the data retention policy for GitHub Coding Agent, specifically regarding how long prompts and any potentially confidential/internal data are stored.
In the official documentation, it states that outside of the IDE (e.g., CLI), prompts and input/output data may be retained for 28 days and are then automatically deleted. Based on that, I assume the Coding Agent may fall under the “outside the IDE” category, but I am not fully sure.
Could you please clarify the following for Business, Enterprise, and Pro plans?
Thank you in advance for your clarification.
Beta Was this translation helpful? Give feedback.
All reactions