Cloudflare Unveils a Firewall Designed to Keep LLMs Safe

Cloudflare announced Firewall for AI and Defensive AI, a set of capabilities designed to protect LLMs, APIs, and email by detecting prompt injections, sensitive data leakage, anomalous traffic, and phishing, while enforcing zero-trust policies. The tools include sensitive data detection with obfuscation, prompt and response validation, and API anomaly detection that learns normal request sequences to flag deviations. The report also highlights the unique security challenges of LLMs compared to traditional applications, noting their non-deterministic behavior and embedded training data, and positions these offerings as proactive defenses against evolving AI-driven threats.