Beyond the ban: A better way to secure generative AI applications

This post introduces AI prompt protection as a new capability within Cloudflare's DLP product, integrated into Cloudflare One. It details the technical components: prompt detection (leveraging undocumented APIs), topic classification (content and intent with predefined profiles for PII, credentials, source code, customer data, financial information, code abuse, and jailbreaks), guardrails for granular policy enforcement based on these classifications, and encrypted logging. The implementation leverages Kivera's operation mapping for API analysis and provides Gateway log filters for AI-specific events. The core technical contribution is the system's ability to inspect and control data flow into and out of web-based GenAI applications based on semantic understanding of prompts and responses.