- August 18, 2026 → GitHub Changelog — date GitHub published enterprise managed settings for Copilot in JetBrains
A platform admin who manages GitHub Copilot for a few hundred engineers can now write one policy file and stop every developer from connecting Copilot to an AI server the security team never approved.
“Administrators can now apply consistent controls for everyone on your enterprise’s Copilot plan,” said GitHub in its August 18 changelog GitHub Changelog.
One file now governs every JetBrains Copilot session
The change reaches Copilot inside JetBrains IDEs such as IntelliJ, PyCharm, and Rider, and it lands in four control areas. Admins can require or block specific Copilot plugins and approve which plugin marketplaces developers may install. They can also keep an allowlist and blocklist of MCP servers, so a developer can no longer point Copilot at an unvetted connector.
Central OpenTelemetry is the third area. Managed settings route Copilot’s telemetry to an approved collector and override anything a developer sets locally, so logs leave the building through one door. The fourth area is permission modes: an admin can disable Bypass Approvals or Autopilot, the modes that let the agent act without asking GitHub Changelog.
Why admins wanted this, and what developers lose
The appeal is obvious for security and compliance teams. A single managed file replaces a hundred local configurations, and managed values win over developer settings every time.
There is a cost. Developers who liked adding their own marketplace or testing an experimental MCP server lose that freedom, and the change can feel like a leash rather than a guardrail. Concerns about autonomy are real: some keys, such as permission modes, cannot be relaxed per team, so a cautious default applies to everyone GitHub Docs.
Where the controls stop
Not every setting is locked. GitHub’s reference shows that plugin and marketplace keys work additively — an enterprise team can add more on top of the baseline — and team-specific overrides are possible for a handful of keys marked “overridable” GitHub Docs.
The precedence order is strict: device policy beats server policy, which beats file-based settings, which beats the user. That hierarchy is governance, not a promise that every risky prompt is caught, and it sits alongside the memory and model controls GitHub has already shipped for JetBrains GitHub Copilot for JetBrains gains memory and Ollama.
For companies weighing a wider Copilot rollout, the question is no longer only whether the assistant can write good code. It is who gets to decide what it is allowed to touch.
