By exploiting how AI coding agents retrieve and verify plugins, researchers were able to execute malicious code even when the agent was told to use a trusted, approved version.

One cloud region, four AI giants, and a preview of what happens if your agents all share the same failure domain.

AI agents used shared infrastructure, like package caches, to break isolation — proving environment signals are no substitute for real authorization checks.

Learn how to seize the power of Python’s dataclasses, editable installs, virtual environments, and new free-threaded build.