Big security + usability upgrade for GitHub Copilot: local sandboxing (on your own machine) is finally usable, tightening what AI agents can do when they run commands. This rollout covers Copilot CLI, the Copilot desktop app, and VS Code sessions using Agent Host. The headline change: when an agent needs to execute commands or access resources, developers and enterprise admins can now restrict permissions—what files it can read/write, whether it can access the network, and whether it can use Git/GitHub CLI credentials. The goal is simple: put agent execution inside a constrained environment so over-permission doesn’t turn into unintended system access. Admins set the rules centrally, which helps prevent accidental misconfigurations by individual users. Under the hood, Copilot’s sandbox uses Microsoft’s eXecution Container (MXC) to map the same policy across Windows, macOS, and Linux isolation mechanisms. Also new in Copilot CLI: model discovery for local Ollama via /model (since 1.0.94-0). It lists models but won’t auto-install or auto-write config—you still confirm and enable. #GitHubCopilot #AISecurity #DevOps #VSCode #Ollama #LocalSandbox
Want to learn more? Visit Explore the world, stay updated on travel insights and international affairs, and discover authentic stories from real life
评论
发表评论