Admins can now disable Copilot and Chat in VS Code with new enterprise controls

Visual Studio Code just got a more locked-down version of AI. Not better AI, exactly — more governable AI. The change is aimed at companies that want Copilot inside the editor, but only on terms the company controls. That matters because VS Code is no longer just sprinkling autocomplete into a text editor. Copilot in VS Code now includes chat, autonomous agents, browser actions, hooks, MCP-connected tools, and extension-provided tools. Once those features start touching code, terminals, and internal systems, admins stop treating them like personal preferences and start treating them like enterprise policy. (code.visualstudio.com) ### What actually changed in VS Code? Microsoft published a new enterprise AI settings page for VS Code that spells out policy controls admins can push to managed devices. Those policies can disable agent mode entirely, turn off hooks, block extension-contributed chat tools, disable browser tools inside chat, and switch off chat plugin integration. On ma(code.visualstudio.com)kspace settings. (code.visualstudio.com) ### Why is that a bigger deal than a toggle? Because these are not cosmetic switches. Agent mode can edit files, run terminal commands, and work through multi-step tasks. Hooks can run shell commands at key points in an agent session. Browser tools let chat agents open and interact with web pages. Extension tools let third-party add-ons expose more capab(code.visualstudio.com)ul autocomplete” and more like a new execution surface. (code.visualstudio.com) ### Can admins really force this on users? Yes — that is the whole point of the enterprise policy system. VS Code says policy values override default, user, and workspace settings. The policies can be deployed through Windows Group Policy, MDM profiles on macOS, and Linux configuration paths, so companies can enforce the same rules across managed fleets instead of asking developers to behave. (code.visualstudio.com) ### Is this the same as blocking Copilot licenses? Not quite. VS Code policy controls govern what the editor can do on the device. GitHub’s own admin controls govern who gets Copilot access and which features or models are allowed at the organization or enterprise level. GitHub already lets enterprise owners disable Copilot for some or all organizations, an(code.visualstudio.com) the new VS Code controls sit one layer lower — inside the client. (docs.github.com) ### So where did the “disable Copilot and Chat” framing come from? It is directionally true, but the official docs are more specific. The new VS Code page clearly documents controls for agents, hooks, browser tools, extension tools, and plugin integration. The material surfaced in the official docs does n(docs.github.com)ncrete: admins can now centrally shut off specific AI behaviors in VS Code itself. (code.visualstudio.com) ### Why now? Because Copilot has expanded fast. The current VS Code docs describe agents that can plan, edit, run, and hand off tasks across sessions, plus support for third-party agents and MCP-based tooling. Once AI inside the editor starts acting more like a junior operator than a suggestion engine, enterprises need the same kind of guardrails they already use for extensions, updates, and network access. (code.visualstudio.com) ### What is the practical effect for developers? If your company manages VS Code, some AI features may simply disappear. Agent mode might be missing. Hooks may do nothing. Browser actions may be unavailable. Chat can still exist in a narrower, safer form — for example, ask or edit flows without autonomous execution. That is the compromise here: keep the useful parts, fence off the risky ones. (code.visualstudio.com) ### Bottom line? VS Code’s AI stack is growing up into enterprise software. And enterprise software gets governed. The new controls do not just make Copilot optional — they make its most powerful behaviors centrally enforceable. (code.visualstudio.com)