GitHub Copilot Training and Your Code Privacy Controls

Whether GitHub Copilot uses your code for model training depends on which Copilot product you use, your account type, and the applicable settings and policies. For Copilot Free, Pro, and Pro+, GitHub’s policy has allowed use of interaction data for model improvement unless you opt out; Copilot Business and Enterprise customer data is not used to train GitHub’s models.

What “training on your code” means

Copilot can process code and surrounding context to generate suggestions. That processing is not the same as adding your code to a training dataset. Training means using data to adjust a model so that it may influence later outputs; inference means using an existing model to produce a response to a current request.

Even when data is not used for training, a service may need to process it to provide features, protect the service, or investigate abuse. The precise data collected and retention periods depend on the product, feature, and current GitHub policy. So “not used for training” should not be read as “never sent to GitHub” or “never retained.”

How Copilot’s controls differ by plan

GitHub’s policies distinguish individual plans from organization plans. Individual users should check the current privacy setting for model improvement in their Copilot settings. GitHub has described an opt-out control for eligible individual plans; turning it off prevents covered interaction data from being used to improve models going forward, subject to the policy’s scope and effective date.

• Copilot Free, Pro, and Pro+: GitHub’s individual-plan policy has included a model-improvement setting. Check your account’s current setting rather than assuming it is off or on.
• Copilot Business and Enterprise: GitHub says customer data from these business offerings is not used to train its models. Organization administrators can also manage access and policies for their members.
• Third-party services: Some Copilot features may involve models or services from providers other than GitHub. Applicable terms and data handling can depend on the feature and configuration.

Policies can change. The GitHub Copilot Trust Center, product documentation, and the privacy controls shown in your account are the best sources for the terms that apply now. If you use Copilot through an employer or school, its organization policy may govern your access.

What data Copilot may process

To respond usefully, Copilot may receive prompts, code snippets, relevant open-file content, and other context made available by the feature you are using. Chat, inline suggestions, code review, and other functions can work differently, so the information sent is not necessarily limited to the line where your cursor sits. Extensions and integrations may have their own data flows.

For example, asking Copilot Chat to explain a function can provide the prompt and selected code as context. An inline completion can use nearby code to predict what comes next. Neither example alone tells you whether the information is used for model training: that depends on the plan, setting, feature, and governing policy.

Never include secrets such as passwords, API keys, private customer data, or confidential source code in a prompt unless your organization has approved the use and the service configuration. A training opt-out is not a substitute for access controls, secret scanning, or following workplace rules.

How to review or change your privacy setting

For an individual account, sign in to GitHub and open your account settings. Find the Copilot section and review the privacy or model-improvement control available there; labels and layout may change. Read the accompanying explanation before changing it, and verify the saved state.

1. Confirm whether you are using an individual plan or an organization-managed Copilot account.
2. For an individual plan, open GitHub account settings and locate Copilot’s privacy or model-improvement options.
3. Set the option according to your preference, then check that the setting was saved.
4. If the account is managed by an organization, ask its GitHub administrator which policy applies and whether members can change the setting.
5. Review the terms for any specific feature or integration that sends code to another service.

Changing a setting generally governs future data use within its stated scope; it may not undo processing that already occurred. Consult GitHub’s current documentation for the setting’s effective date and exceptions.

What the controls do not guarantee

A privacy control is one part of data governance, not a guarantee that generated code is private, unique, correct, or free of similar text elsewhere. Copilot’s suggestions can resemble publicly available code. Review suggestions before using them, especially for licensing, security, correctness, and fit with your project.

Likewise, blocking model training does not necessarily disable Copilot, stop the service from processing context to answer a request, or control every retention, telemetry, and safety practice. Those are separate questions answered by the terms and documentation for the plan and feature.

For company code, the practical choice is usually to use an organization-approved plan and configuration, keep sensitive material out of prompts unless permitted, and confirm the policy with the administrator. For personal projects, inspect the current account setting and decide whether the feature’s data handling meets your expectations.

Frequently Asked Questions

Does GitHub Copilot train on private repositories?
Private repository status does not by itself determine training use. The applicable Copilot plan, privacy setting, feature, and policy matter. Business and Enterprise customer data is not used to train GitHub’s models under GitHub’s stated policy; individual users should check their current setting.

Does turning off model improvement stop Copilot from seeing my code?
No. Copilot may still process relevant code and prompts to provide a feature. The control concerns covered use for model improvement, not whether context is transmitted for service operation.

Is Copilot Business safer for company code?
It has organization-oriented controls and GitHub states Business and Enterprise customer data is not used to train its models. It is not a blanket guarantee against exposure: configure access appropriately and follow your organization’s data rules.

Can I change the setting later?
Check the Copilot privacy controls in your account. A changed preference may apply only to data handled after it takes effect; consult the current policy for details rather than assuming earlier data is removed.

Should I paste an API key or confidential code into Copilot Chat?
No, not unless your organization explicitly permits it and the relevant configuration is approved. Training settings do not make secrets safe to share.

Related

• Continuous Delivery vs Continuous Deployment: Key Differences
• Install TWRP on Huawei Honor 4X: Full Guide
• How to Choose a Gaming Mouse for Grip & Games