arXiv cs.CR·19d agoRevoked but Still Authoritative: An Empirical Study of Revocation Enforcement in Agent-Memory Systems#agent-memory#guard#llm-agentsAI safety & security