TrojAI

TrojAI is an enterprise AI security platform for organizations deploying AI models, applications, and agentic workflows in production, including regulated industries and MSSPs managing AI risk.

Key capabilities include:

• Build-time model and agent risk assessment (TrojAI Detect) with automated red teaming and 100+ security tests to surface vulnerabilities, unsafe behaviors, and remediation guidance before deployment
• Runtime protection (TrojAI Defend) that acts as an AI application and agent firewall, defending against prompt injection, jailbreaking, data poisoning, and sensitive data leakage
• Agent-Led AI Red Teaming using autonomous, multi-turn attacks mapped to security frameworks such as OWASP, MITRE, and NIST, plus Agent Runtime Intelligence for visibility into execution traces, tool usage, and memory access
• Integrations for enterprise governance, including ChatGPT Enterprise API policy enforcement, compliance logging, and risk detection

Pricing: Contact for pricing.

What makes it notable: It combines pre-deployment automated red teaming with in-production controls and observability for agent behavior, including protections for coding agents (for example, detecting exposed secrets and blocking injection attempts), with support for compliance needs such as GDPR and HIPAA.