Building an evidence-grounded agentic security operations harness on Cloudflare
Cloudflare has integrated a new agentic security operations framework into Managed Defense that utilizes Workers for automated evidence collection. The system decouples deterministic data gathering from LLM-based inference to provide grounded security recommendations.
Verified State Diff
Impact & Verification Analysis
Cloudflare Managed Defense customers and security operations center (SOC) analysts.
This represents a shift toward agentic AI architectures in security, prioritizing verifiable evidence over raw model output to increase the reliability of automated threat detection and response.
Full Fact Overview
The architecture leverages Cloudflare Workers to execute deterministic data collection tasks across the global network telemetry layer before passing structured evidence to AI models for inference. By separating the data retrieval process from the reasoning engine, Cloudflare aims to reduce hallucinations and improve the accuracy of security alerts handled by Managed Defense Analysts.