CPython Sandbox Escape Flaw Threatens AI Supply Chain

RunSybil researchers demonstrated a CPython memory-safety flaw enabling sandbox escapes in AI inference servers like vLLM.

A memory-safety flaw in CPython was found by runsybil.com and submitted as GitHub pull request #155761. It threatens the AI supply chain. Researchers demonstrated exploiting the flaw in a backdoored Qwen2.5-0.5B-Instruct model to escape sandboxes and execute code in vLLM inference servers. The issue affects providers enabling bring-your-own-model uploads, potentially exposing infrastructure for multiple customers. The Python Security Response Team disputed that it qualifies as a CPython security vulnerability, stating sandbox escapes fall outside their threat model and belong with sandbox maintainers. The researchers submitted a patch that merged into the main branch and was backported to Python 3.13, 3.14, and 3.15.