GitHub Copilot Telemetry in VS Code: Settings, Training and OTel

To reduce telemetry in VS Code, start with telemetry.telemetryLevel. But do not treat it as an offline switch or a complete Copilot privacy policy. Editor diagnostics, extension telemetry, AI request processing, model-training preferences and an organization’s OpenTelemetry export are different controls.

This guide separates those controls so you do not rely on a setting that merely looks plausible in JSON.

What the VS Code Telemetry Setting Controls

VS Code documents four levels: all includes usage, error and crash information; error excludes usage data; crash retains crash reporting; and off disables the editor’s telemetry reporting going forward. It does not retract data already sent. Disabling usage telemetry also disables experimentation, so some new features may arrive later. See VS Code’s telemetry documentation.

The same documentation warns that extensions can have their own collection. Extension authors are instructed to respect VS Code’s telemetry API, but that guidance is not evidence that a single preference blocks every installed extension or external service.

Turn Off Editor Telemetry

1. Open VS Code Settings and search for telemetry.telemetryLevel.
2. Set Telemetry: Telemetry Level to off. On a managed device, ask your administrator if a policy prevents changes.
3. Alternatively, use Preferences: Open User Settings (JSON) and add the setting below to the existing object, preserving the file’s other settings and valid commas.

"telemetry.telemetryLevel": "off"

Do not add github.copilot.advanced.telemetry.optOut: it is not a documented current Copilot control. An unknown key sitting in settings.json does not enforce a privacy restriction. Use the current telemetry-level control rather than old separate crash-reporter examples.

Review Copilot’s Model-Training Preference Separately

For individual Copilot Free, Pro, Pro+ and Max plans, GitHub’s current policy permits use of interactions for model training unless you opt out. To opt out, open your GitHub profile menu, choose Copilot settings, and set Allow GitHub to use my data for AI model training to Disabled.

This training preference is not the VS Code diagnostic setting. GitHub states that Business and Enterprise customer data is not used for model training without customer authorization under its applicable data-protection agreement; the individual training control therefore does not appear for those seats. Confirm the signed-in account and plan rather than assuming the missing control is an error. See GitHub’s current policy controls.

Opting out of training does not mean no request data reaches the service when you use a cloud AI feature.

Check OpenTelemetry and Organization Monitoring

Copilot Chat can separately export agent traces, metrics and events to an OpenTelemetry collector. This export is off by default. Current settings include github.copilot.chat.otel.enabled and a separate content-capture control. If export was enabled for diagnostics, review its destination and what it records before leaving it enabled.

On managed devices, enterprise telemetry settings can override user preferences. Do not attempt to bypass required organizational monitoring; ask the administrator which policy applies. Also distinguish remote export from local debug information: disabling export does not guarantee that no local session or debug records exist.

The OpenTelemetry guide describes export, content capture and managed settings; the AI settings reference lists current controls.

Verify the Configuration Without Claiming “No Data Leaves”

Reopen Settings to confirm the effective telemetry level. Review the installed extension’s current documentation and the account-level training preference. If your organization uses a collector, verify its configuration with the administrator rather than inferring it from an empty local log.

Developer: Show Telemetry can expose VS Code telemetry events for investigation, but a quiet output pane is not a network audit. Diagnostic exports can themselves contain sensitive material; inspect and redact them before sharing.

If your requirement is that proprietary source must never leave a machine, telemetry opt-out alone is insufficient. Use an approved offline workflow or disable the relevant cloud features according to your organization’s policy. Do not assume a working cloud Copilot session is local inference because telemetry is off.

Official documentation checked: October 14, 2026.