Glossary
LangChain Deep Agents: What the Architecture Means for AI Agent Security
Deep Agents don’t add new AI capability, they change the architecture. Planning, memory, filesystem storage, and subagents make agents more powerful but also expand the security surface. Stored content becomes a prompt injection risk, memory spans sessions, and agent handoffs create trust gaps. It’s more observable, but only secure if teams enforce strong controls around tools, storage, and monitoring.
