CISA added CVE-2026-40217 to its exploited-vulnerability catalog on November 18 alongside flaws in Apache Axis, REDCap and Synology. For AI operators, the LiteLLM issue exposes the danger of treating a guardrail-testing feature as a secure code sandbox. The maintainer explains that POST /guardrails/test_custom_code executes supplied Python and that bytecode manipulation can escape its original restrictions. In default configurations the endpoint requires a proxy-admin credential; this advisory does not describe ordinary inference access alone as sufficient. Code runs with the proxy process's privileges, which were root in the default container described by the researchers. The maintainer's version table lists 1.83.10 as patched, while its narrative explicitly directs users to 1.83.11 or later and describes replacing the sandbox with RestrictedPython. That discrepancy should remain visible in patch verification. CISA's inclusion establishes observed exploitation; it does not prove that every reachable gateway was compromised or that a software update removes already-established access.
LiteLLM enters CISA KEV: restrict guardrail testing and verify the deployed fix
Related research
More curated notes connected through Agent Security and AI Engineering.
OWASP Top 10 for Agentic Applications for 2026
OWASP's community guide organizes agentic-system risk into ten categories, including goal hijacking, tool misuse, identity and privilege abuse, memory poisoning, insecure inter-agent communication, cascading failures, and rogue-agent behavior. It provides a shared taxonomy and mitigation starting point rather than a certification checklist or evidence that a deployed system is secure.
A five-step roadmap to closing the AI evaluation gap
The roadmap addresses evaluation results that overstate real-world performance or fail to transfer across deployment contexts. Its five steps balance standardized and local tests, evaluate throughout the lifecycle, build qualified assurance and communication capacity, tailor tests to each value-chain actor and technology, and use a coordinated, trusted process for updating methods.
FinBot CTF Is Live: A Hands-On Companion to the OWASP GenAI Security Project
OWASP FinBot is a hands-on agentic-security CTF built around a simulated multi-agent financial-services platform with real tool access. Its challenges cover prompt injection, tool misuse, policy bypass, data exfiltration, privilege escalation, remote code execution, shared context, and compromised MCP servers.