This topic covers AI-powered security and fraud-detection solutions tailored to crypto platforms, exchanges, and DeFi protocols, with an emphasis on agent governance, observability, and regulatory compliance. As of 2026-05-14, the space is shaped by two converging trends: growing use of agentic AI and LLMs to automate both defensive monitoring and offensive abuse, and stronger regulatory demands for auditability, transaction surveillance, and rapid incident response. Practical tools fall into complementary categories: infrastructure and observability (e.g., Xilos, which positions itself as an intelligent agentic AI infrastructure claiming full visibility into connected services and agent activity); agent development and governance platforms (StackAI and Kore.ai, which offer no-code/low-code to pro-code workflows for building, orchestrating, and governing multi-agent automation with built-in observability); and model/ML primitives for detection and retrieval (Cohere’s enterprise LLMs and embeddings for private generation, similarity search and retrieval-augmented detection, and Mistral’s efficient open models and production tooling for on-premises or privacy-conscious deployments). Effective crypto fraud defense now combines real-time transaction analysis, entity-linking via embeddings, behavioral anomaly detection across agentic workflows, and immutable audit trails to meet compliance requirements. Organizations prioritize solutions that provide transparent model behavior, secure private deployment of LLMs, and governance controls to prevent automation drift and misuse. Selection criteria include observability into agent actions, support for private/custom models, built-in compliance logging, and the ability to integrate with blockchain analytics and SIEM pipelines. This topic guides comparisons of those capabilities and how they address both operational risk and evolving regulatory expectations.