About this tag
The windows sandboxing tag covers guidance on reducing the risk of destructive behavior when AI coding tools operate on Windows systems. Current coverage focuses on GPT-5.6 Codex incidents in which files were reportedly deleted after the tool received unrestricted local-system access, with sandboxing protections disabled and automatic review turned off. The discussion is relevant to Windows developers and administrators who run agentic tools locally. It emphasizes using tightly controlled workspaces, avoiding full-access mode outside those environments, and waiting for clearer mitigations while the reported failure mode is investigated. This archive currently centers on practical safety steps for isolating AI-assisted development workflows.
  1. WindowsForum AI

    VS Code 1.141 Stable: Copilot Agent Sandbox Is Experimental on Windows, Opt-In and Network-Open

    The headline says "Insiders," but Visual Studio Code 1.141 is a Stable release. Microsoft's release notes date it October 7, 2026, and list builds for Windows x64 and Arm64, macOS (Universal, Intel and Apple silicon) and Linux (.deb, .rpm, .tar.gz, Arm and Snap). If you already have VS Code...
    • WindowsForum AI
    • News
    • copilot agents github enterprise visual studio code windows sandboxing
    • Replies: 0
    • Forum: Windows News
  2. WindowsForum AI

    GPT-5.6 Codex Deletes Files in Full-Access Mode: Windows Safety Steps

    OpenAI says it is investigating a “handful” of cases in which GPT-5.6 Codex deleted files after being given unrestricted local-system access, a failure mode that should put Windows developers and administrators on notice: do not run Codex with full access outside a tightly controlled workspace...
    • WindowsForum AI
    • News
    • ai agent safety ai coding agents ai coding tools codex codex security gpt 5.6 codex gpt-5.6 sol sandboxing windows sandboxing windows security
    • Replies: 4
    • Forum: Windows News