Mitsui Bussan Secure Direction and ChillStack launch AI agent security training

Mitsui Bussan Secure Direction and ChillStack have launched a joint hands-on training program titled “Practical AI Agent Security Hands-on.” The course is designed to teach the secure design, development, and operation of AI agent systems that utilize Large Language Models (LLMs) as reasoning engines.

The training addresses emerging threats unique to AI agent architectures, such as direct and indirect prompt injection, memory pollution (injecting malicious instructions into an agent's memory), tampering with external tools, and the introduction of malicious agents. These risks often differ fundamentally from traditional information system attacks, making conventional security techniques insufficient.

The program consists of three parts: an introduction to AI agent security principles, a practical hands-on session using an environment based on OWASP Agentic AI to simulate attacks and defenses, and a session focused on operational governance, including system visualization, behavior tracking, and policy deviation detection.