Elastic Advances LLM Security with Standardized Fields and Integrations

This Elastic Security Labs report outlines a standardized approach to securing LLM ecosystems, detailing an AWS Bedrock integration that maps model invocation logs to Elastic Common Schema (ECS), proposed gen_ai fields, guardrail handling, and example detection rules and queries (sensitive-content refusals, high token usage/DoS, latency anomalies, malicious model uploads, and suspicious external communications) to help organizations audit, detect, and respond to misuse of generative AI.