Block unsafe prompts targeting your LLM endpoints with Firewall for AI

Cloudflare announced an expansion of its Firewall for AI to include unsafe content moderation using Llama Guard, enabling detection and blocking of harmful or sensitive LLM prompts (e.g., hate, violence, sexual content) at the network edge with unified analytics and policy enforcement. The system leverages an asynchronous, scalable architecture on Workers AI to minimize latency, exposes new analytics fields for unsafe topic insights, and allows custom rules to log or block categories, with a roadmap including prompt-injection/jailbreak detection and response handling; the feature is available in beta.