TrojAI Defend is an AI firewall that provides runtime protection for AI models, applications, and agents in production environments. The product monitors, alerts, blocks, redacts, and logs threats to AI systems through real-time filtering of inputs and outputs. The solution addresses multiple AI attack vectors including prompt injection, jailbreaking, unbounded consumption, sensitive information disclosure, toxic and harmful content, improper output handling, vector and embedding weaknesses, system prompt leakage, and misinformation. It implements content moderation to sanitize model prompts and responses. TrojAI Defend includes an AI-powered rules engine that supports both out-of-the-box and customized rules for detection and prevention. The platform scales to filter more than 10 million tokens per second for enterprise deployments. Browser extensions enable employees to use third-party GenAI and copilot applications while protecting against data loss of PII, IP, and proprietary information. The product provides reporting and visibility through integrations with SIEM, SOAR, and ticketing platforms. It maps to AI security standards including OWASP, MITRE, and NIST for compliance purposes. The solution operates as a firewall layer that can be deployed across multiple AI applications and use cases.