Coding · Source-backed update

Copilot separates identity from content capture

Enterprise settings can attribute activity to an account without enabling prompt or tool-content capture.

Edited by Clinton FeyisitanEvent Published

Review scope: GitHub’s announcement, managed-settings reference and linked VS Code monitoring documentation, read 13 October 2026. No client configuration, telemetry export or native workflow was executed.

What changed

GitHub’s 12 October announcement separates identity attribution from five content-capture categories: prompts, responses, tool arguments, tool output and policy details. Each category defaults to off. Identity capture records the authenticated account, operating-system account and machine name; it remains separate from the existing pseudonymous identifier.

The announcement lists Copilot CLI, the GitHub Copilot app and VS Code Agent Host sessions. Attribution and content are separate controls in that announced design. The VS Code documentation has conflicting scope statements, outlined below.

The VS Code distinction matters

The managed-settings reference distinguishes Agent Host from VS Code’s local agent harness. The local harness supports identity and the older combined content setting, but does not apply the five content switches independently.

According to GitHub’s managed-settings reference, explicit managed values override local settings, SDK configuration and environment variables. The legacy captureContent switch controls the five content classes together, never identity. A specific capture-class setting overrides that shorthand. When several managed sources define telemetry, the highest-precedence source supplies the whole object; lower-precedence objects are not merged.

Where the documentation disagrees

The linked VS Code monitoring guide, read the same day, limits identity attributes to the Local harness and says Agent Host telemetry omits them. It also allows OpenTelemetry environment variables to override managed values in the Copilot Chat extension. Those narrower statements do not match the broad support and precedence wording above.

Fewertools has not reconciled that discrepancy through a native test. Treat the announcement as the announced scope, and verify the exact client, harness and version before relying on attribution or enforcement.

Keep a reproducible record

The VS Code monitoring guide documents OpenTelemetry collection. For a scoped trial, record the client, harness, effective settings and collector alongside the task. The coding task worksheet provides a place to keep those conditions.

This documentation check establishes neither successful delivery to a collector nor a code-quality result. See the Copilot profile and Cursor/Copilot comparison for the separately dated feature, price and editorial scopes.

Sources and limits

Source review only. No account settings were changed, no telemetry was collected and no pricing or editorial scores were refreshed.

More dated tool changes · How Fewertools records evidence