GitHub Copilot has moved Assisted Permissions from preview to general availability for the CLI, the desktop application, and the VS Code extension. When the tool needs to execute a shell command, read or write a file path, or open a URL, an integrated large‑language model evaluates the request and automatically allows actions it deems clearly safe, while flagging ambiguous cases for user approval.
How to Enable Assisted Permissions
Each client has a distinct activation path:
- CLI: run
copilot /permissionsand choose the Assisted mode. - Desktop app: open the mode picker, select Tool permissions, then pick Assisted.
- VS Code: open the permissions picker from the extension UI and switch to Assisted permissions.
Enterprise Controls and Policy Interaction
Administrators can manage Assisted Permissions through enterprise‑managed settings. Options include:
- Blocking the “Allow All” mode while keeping Assisted enabled.
- Disabling Assisted Permissions entirely.
- Defining the default permission mode for new Copilot sessions.
Developers may still select any mode that the organization’s policy permits. If the configured default is unavailable or prohibited, the session falls back to manual approval for every request.
Operational and Security Implications
From an implementation perspective, the LLM‑based evaluator introduces a new decision point between the user and the underlying system. Engineers should consider:
- Workflow latency: Requests that require manual approval add a pause, which can affect interactive development or automated scripts.
- Risk reduction: Automatically approved “clearly safe” actions reduce the chance of accidental privileged operations, but the classification criteria are not exposed, so teams must treat the model’s judgment as advisory.
- Policy alignment: Enterprise settings can enforce a consistent default, helping to standardize how permissions are granted across teams.
Related CloudNinjas coverage: AI engineering.
What This Means For Practitioners
Adopt the Assisted mode where you want a balance between convenience and oversight, and verify that your organization’s policy settings reflect the desired default. Monitor the frequency of manual approvals to gauge the LLM’s classification accuracy and adjust policies if the approval burden becomes excessive. Finally, incorporate the activation steps into onboarding documentation so new developers understand how to switch modes and what to expect when a request is flagged for review.
