CVE-2026-70335

Think of Visual Studio Code as a command “switchboard” that helps run actions on your computer. This flaw can be tricked so that, when a user interacts with a malicious Copilot/CODE workflow, it runs attacker-chosen commands—like replacing the switchboard instructions with a malicious one. That can allow an attacker to take over what the app is allowed to do on that machine.

This matters if your business uses Visual Studio Code, especially with GitHub Copilot, on machines where an attacker could get local access (for example, through a shared device, compromised workstation, or malicious content that influences user interaction). The risk is mainly for scenarios where a local attacker can induce user interaction with a malicious payload inside Copilot or Visual Studio Code. If the attacker can’t reach your machine locally or can’t cause the user interaction, the exposure is much lower.

This is RED because the flaw is about executing arbitrary operating system commands and is triggered by user interaction, and the fix is already available. Even though no clear, dated incidents are reported publicly for this CVE, the involvement of a well-known target (Visual Studio Code) and recent actor attention means you should treat it as active patch-priority work.