On September 9, 2026, GitHub announced GitHub Copilot managed permissions for agent operations. Copilot Business and Enterprise administrators can centrally block an action, require human approval, or permit it without a prompt. The controls cover commands, file access and edits, and network domains, according to GitHub’s announcement.
What changed
The granular controls are generally available in the Copilot app, Copilot CLI, and VS Code sessions using Agent Host. GitHub says local preferences and saved approvals cannot relax managed restrictions. Enterprises can also specialize policies by team. This release adds operation-level controls beyond the managed-settings deployment options QATechTools covered in July.
The official permissions reference defines the order as deny, then ask, then allow. A managed ask rule needs a new approval each time the operation is requested. When multiple managed sources define allowlists, only their intersection is allowed. With an applicable permission rule or declared allowlist, unmatched supported operations require approval.
Why this matters for QA engineers
These boundaries become part of the environment in which an agent generates tests or investigates failures. Our recommendation is to validate permission behavior alongside test results before expanding an agent pilot.
Test three outcomes: use harmless fixtures to confirm that a blocked file stays unread, a permitted test command runs, and an approval-gated action waits for a person. Repeat the gated action to check that earlier approval does not silently authorize the next request.
Test conflicts: introduce overlapping deny and allow rules in a disposable environment, then verify the denial wins. Record the client version, session type, effective policy and observed operation outcome so a later regression is reproducible.
For VS Code, confirm the session uses Agent Host before expecting these granular rules to apply. Continue reviewing generated assertions and regression evidence: permission checks establish which actions may run, while functional tests establish whether the resulting code behaves correctly.
Sources: GitHub announcement dated September 9, 2026; GitHub documentation checked September 10, 2026. QA rollout recommendations are QATechTools analysis.