https://miasyster.github.io/en/tags/sandbox/Recent content in Sandbox on AI + Quant EngineeringHugoenFri, 03 Apr 2026 00:00:00 +0000https://miasyster.github.io/en/posts/sandbox-defense-in-depth/Fri, 03 Apr 2026 00:00:00 +0000https://miasyster.github.io/en/posts/sandbox-defense-in-depth/AI-generated code must be executed — otherwise it’s just text. But execution means risk. I didn’t choose container isolation or RestrictedPython. Instead I designed a three-layer defense: reject dangerous structures at compile time via AST, replace the entire builtins at runtime, and enforce OS-level resource limits as a backstop. Each layer handles a different class of risk. Overlapping but not redundant.