TrojAI Detect is an AI security platform designed to identify vulnerabilities and weaknesses in AI models during the build phase. The product focuses on testing and monitoring AI model behavior to prevent adversarial attacks, data leakage, and other security risks before deployment. The platform addresses multiple AI security concerns including prompt injection attacks, jailbreaking attempts, PII leakage, data loss prevention, and toxic content generation. It provides automated detection of potential problems in AI models across different environments and cloud platforms. TrojAI Detect includes capabilities for mapping AI models and applications to security standards and frameworks such as OWASP Top 10, MITRE, and NIST. The platform supports public, private, and custom AI models across multiple cloud environments. The product features a customizable risk engine with pre-built and custom policies, designed to handle enterprise-level production workloads. It offers flexible deployment options with self-hosted capabilities to maintain data security within customer environments. TrojAI Detect includes blocking rules for various threat categories including PII detection, named entity recognition, toxicity detection, DoS attack detection, source code detection, and confidential data loss prevention. The platform provides dashboards for monitoring firewall events and tracking blocked threats in real-time.