This session will introduce Chagu, a novel protocol currently under pre-patent review, designed to defend generative AI systems using graph-based techniques. Talex Maxim will demonstrate how Neo4j empowers Chagu to trace prompt execution paths, visualize AI state drift, and detect anomalies in distributed, multi-agent environments.

You will learn how to model LLM interactions, suspicious flows, and behavior shifts using graph structures, making use of Cypher queries to identify threats in real-time. This session will provide an actionable blueprint to help you secure agentic systems by combining graph data models, Cypher-powered detection, and open-source anomaly defense strategies.

Attendees will walk away with code snippets, graph modeling ideas, and inspiration for combining Neo4j with LLM pipelines, security layers, and data-driven observability.