00:02
2026-05-28
arxiv.org
ai-safety
Agent Security Is a Systems Problem
Researchers argue that AI agent security must be treated as a systems-level problem, with the AI model itself considered an untrusted component and security invariants enforced at the system level rat…