Take control of public AI application security with Cloudflare's Firewall for AI
Mar 19, 2025Radwa Radwan, Clement Bertier, Zhiyuan Zheng
Summary
This post introduces Firewall for AI, an inline security solution for LLM applications. It details the technical approach to discovering LLM endpoints by analyzing API path signatures, server-sent events, and response bitrate heuristics. It also describes the detection mechanism using regexes and a NER model (Presidio) deployed on Workers AI. The integration with the WAF and Security Analytics for real-time mitigation and visibility is also explained.
More from the Website Security & Threat Management trail
- We tested our own WAF with frontier AI models. Here’s what we found
- When scanners miss the attack: how Cloudflare Client-Side Security protects storefronts
- Introducing automatic remediation policies with Cloudflare CASB
- Introducing context-aware vulnerability discovery and remediation with Cloudflare Managed Defense and OpenAI Daybreak Models
- Introducing Adaptive Intelligence: undermining the economics of every bot attack