Time Table

With the appearance of agent protocols such as MCP (Model Context Protocol) and A2A (Agent-to-Agent), Multi-Agent Systems (MAS) are drawing significant attention. MAS connect multiple AI agents and external tools over networks, enabling them to autonomously execute complex processes. They are expected to become increasingly integrated into society in the coming years.

The role that the traditional Web has played—a space where humans access information and use services through a web browser—is being taken over by MAS, evolving into a new form of the Web called the “Agentic Web,” where AI agents interact directly with each other to deliver the most relevant information and services to humans. In the Agentic Web, critical decisions and transactions are carried out by AI agents without human involvement. This shift introduces new threat models in addition to conventional Web-security concerns. Examples include tampering or impersonation in agent-to-agent communications, the injection of malicious tools, and data poisoning leading to faulty decisions—risks that cannot be fully mitigated by browser-centric security measures alone.

This session will explain the following three perspectives in an accessible manner and highlight the security challenges that organizations will face during this transformation, as well as how to prepare for them:

• The architecture and social impact of the Agentic Web
• Emerging threat scenarios revealed by the differences from the traditional Web
• Core principles and defense strategies for securely leveraging MAS

This talk is ideal for anyone seeking a comprehensive introduction to Agentic Web Security as a first step toward preparing for the next generation of the Web.

Speakers:
Isao Takaesu（高江洲　勲）Mitsui Bussan Secure Directions, Inc. Senior Professional, Product & Solution Division