NICT and MITRE to Launch AI Security Competition in the United States

The National Institute of Information and Communications Technology (NICT, President: OHNO Hideo, Ph.D.) promotes research and development (R&D) and international collaboration in AI security through the Center for Research on AI Security and Technology Evolution (CREATE) within its Cybersecurity Research Institute. Building on these efforts, NICT and The MITRE Corporation (MITRE) will jointly launch an AI security competition called "AI Breach Challenge."

The inaugural event, "AI Breach Challenge 2026: Email Agent Security," will be held on December 16-17, 2026, in the Washington, D.C. area. The competition will focus on AI agents that support email operations. In an isolated and controlled environment provided by the organizers, participants will explore vulnerabilities in AI agents and test attack techniques in practice. By combining their expertise, NICT and MITRE aim to use the findings from the competition to advance future R&D on security evaluation and defense technologies for AI agents.

NICT established CREATE in February 2025 to advance cybersecurity through AI and to develop security technologies for the safe and secure use of AI, in collaboration with research institutions and companies in Japan and abroad. MITRE develops and maintains MITRE ATLAS™, a knowledge base that organizes adversarial tactics and techniques targeting AI systems, and works to improve the understanding of AI security threats and strengthen countermeasures. Drawing on their respective strengths, NICT and MITRE are pursuing collaborative activities in AI security. AI agent security is one of the key topics in this collaboration.

AI agents are increasingly being used as workplace assistants that help users manage email and perform other tasks by connecting to business information and tools. Unlike conventional generative AI, which primarily generates text and other content in response to user instructions, AI agents can interact with external systems and tools, enabling them to carry out a range of actions on a user's behalf. These capabilities can improve productivity, but they may also create new security risks. For example, malicious input could cause an AI agent to perform unintended actions or leak sensitive information to unauthorized parties. As the use of AI agents expands, it is important to understand what attacks are possible, what vulnerabilities and failure patterns may arise, and how effectively existing safeguards withstand real attacks.

Against this background, NICT and MITRE will jointly launch AI Breach Challenge as a concrete initiative arising from their collaboration. The competition will examine AI agent weaknesses from an attacker's perspective and apply the findings to R&D on security evaluation and defense technologies.

AI Breach Challenge will examine security issues across a range of AI agent use cases. The inaugural competition will focus on AI agents that support email operations in a workplace environment, including reading and writing emails and interacting with related tools. Participants will tackle a set of predefined challenges in a competition environment provided by the organizers, exploring vulnerabilities in AI agents and testing attack methods.

The competition will assess AI agent security from an attacker's perspective. Participants will test different approaches, including attempts to cause unintended actions or gain access to protected information. The findings will help identify weaknesses in AI agents and inform future R&D on security evaluation and defense technologies.

As part of the research collaboration between NICT and MITRE, this competition is intended not only as a test of technical skills, but also as a practical research opportunity to better understand security threats to AI agents. By analyzing participants' attack approaches and the outcomes, NICT and MITRE aim to gain knowledge in the following areas:

NICT and MITRE will analyze the findings from the competition and apply them to future R&D on security evaluation and defense technologies for AI agents, aiming to generate research outcomes through their collaboration. These efforts will also strengthen CREATE's international partnerships in AI security, particularly its collaboration with MITRE. NICT and MITRE plan to continue AI Breach Challenge, exploring security challenges across a wider range of AI agent use cases, and contributing to the advancement of AI security research through U.S.-Japan collaboration.

The National Institute of Information and Communications Technology (NICT) is Japan’s sole national research and development (R&D) agency specializing in information and communications technology (ICT), promoting ICT R&D to support economic growth and a safe and secure society. In February 2025, NICT established the Center for Research on AI Security and Technology Evolution (CREATE), which collaborates with research institutions and industry partners in Japan and abroad to advance cybersecurity through AI and develop technologies for the safe and secure use of AI.

MITRE is a not-for-profit organization that operates federally funded research and development centers and works across government, industry, and academia to address challenges of national importance. MITRE developed ATLAS™, (Adversarial Threat Landscape for Artificial-Intelligence Systems), a knowledge base documenting adversarial tactics, techniques, and case studies involving artificial intelligence and machine learning systems.