Notice of Another Star LLC Winning the GENIAC-PRIZE Future Vision Award (Special Award)
Another Star LLC has won the 'Future Vision Award' at the NEDO-sponsored GENIAC-PRIZE for developing an open-source security platform that ensures safe communication between AI agents.
📋 Article Processing Timeline
- 📰 Published: March 31, 2026 at 19:00
- 🔍 Collected: April 1, 2026 at 13:39 (18h 39m after Published)
- 🤖 AI Analyzed: April 22, 2026 at 09:39 (500h 0m after Collected)
Another Star LLC (Headquarters: Minato-ku, Tokyo, Representative Member: Naoya Yasuda, hereinafter 'our company'), which develops a security platform for AI agent linkage, is pleased to announce that it has received the Special Award 'Future Vision Award' (Prize: 10 million yen) in Category 03 'Exploration of risks and development of risk mitigation technologies to ensure the safety of generative AI' at the 'NEDO Prize-based Program GENIAC-PRIZE' sponsored by the New Energy and Industrial Technology Development Organization (NEDO). The Future Vision Award is a special award that evaluates the novelty and future potential of the proposal, and only three companies, including ours, have won this award.
The GENIAC-PRIZE is a prize-based research and development contest established as part of 'GENIAC', a generative AI social implementation support project promoted by the Ministry of Economy, Trade and Industry (METI) and NEDO. It is a program that widely solicits applications aimed at solving social issues using generative AI services and awards prizes (totaling approximately 800 million yen) based on the results. The award ceremony was held on March 24, 2026, at Kanda Myojin Hall (Chiyoda-ku, Tokyo).
Background and Process Leading to the Award
AI agents are rapidly evolving from standalone AIs into entities that autonomously collaborate with multiple external AIs to accomplish tasks. As the standardization of Google's A2A Protocol (Agent-to-Agent Protocol) progresses, an era where AI agents communicate directly without human verification is becoming a reality.
The cybersecurity market for AI agents is predicted to grow at a CAGR of 34.4%, from approximately $22.6 billion (about 3.4 trillion yen) in 2024 to approximately $322.4 billion (about 48.4 trillion yen) by 2033 (according to Grand View Research). However, in this rapidly growing market, most existing security solutions target endpoint or network layer defense, and a 'protocol-level security foundation' that ensures the reliability of communication between AI agents itself has not been sufficiently established either domestically or internationally.
With this structural change, new risks are emerging that conventional security cannot handle. Focusing on this blank space, we identified two risks: (1) the uncertainty of the authenticity and reliability of external AI agents, and (2) the hijacking of AI agents through indirect prompt injection (malicious data being executed as commands). Based on these, we proposed the concept of a 'trust layer required in the era where AI agents communicate with each other'.
In the screening process, our concept of a self-improving loop was highly evaluated from the perspective of novelty and future potential. This loop not only monitors communication between AI agents to detect and block anomalies but also feeds back the detected anomalies to a trust score on the agent store, continuously improving the safety of agent cooperation. Furthermore, through verification using sample agents and face-to-face demonstrations, a series of operations from detecting and blocking anomalies to reflecting them in the trust score were confirmed.
Developed Platform: 'Domestic OSS Secure AI Agent Linkage Platform'
The platform we developed is an unprecedented infrastructure, both domestically and internationally, that provides a trust layer for linkage between AI agents. It consists of a multi-layered defense structure centered around two core components: an 'Agent Store' and a 'Broker Agent'.
The GENIAC-PRIZE is a prize-based research and development contest established as part of 'GENIAC', a generative AI social implementation support project promoted by the Ministry of Economy, Trade and Industry (METI) and NEDO. It is a program that widely solicits applications aimed at solving social issues using generative AI services and awards prizes (totaling approximately 800 million yen) based on the results. The award ceremony was held on March 24, 2026, at Kanda Myojin Hall (Chiyoda-ku, Tokyo).
Background and Process Leading to the Award
AI agents are rapidly evolving from standalone AIs into entities that autonomously collaborate with multiple external AIs to accomplish tasks. As the standardization of Google's A2A Protocol (Agent-to-Agent Protocol) progresses, an era where AI agents communicate directly without human verification is becoming a reality.
The cybersecurity market for AI agents is predicted to grow at a CAGR of 34.4%, from approximately $22.6 billion (about 3.4 trillion yen) in 2024 to approximately $322.4 billion (about 48.4 trillion yen) by 2033 (according to Grand View Research). However, in this rapidly growing market, most existing security solutions target endpoint or network layer defense, and a 'protocol-level security foundation' that ensures the reliability of communication between AI agents itself has not been sufficiently established either domestically or internationally.
With this structural change, new risks are emerging that conventional security cannot handle. Focusing on this blank space, we identified two risks: (1) the uncertainty of the authenticity and reliability of external AI agents, and (2) the hijacking of AI agents through indirect prompt injection (malicious data being executed as commands). Based on these, we proposed the concept of a 'trust layer required in the era where AI agents communicate with each other'.
In the screening process, our concept of a self-improving loop was highly evaluated from the perspective of novelty and future potential. This loop not only monitors communication between AI agents to detect and block anomalies but also feeds back the detected anomalies to a trust score on the agent store, continuously improving the safety of agent cooperation. Furthermore, through verification using sample agents and face-to-face demonstrations, a series of operations from detecting and blocking anomalies to reflecting them in the trust score were confirmed.
Developed Platform: 'Domestic OSS Secure AI Agent Linkage Platform'
The platform we developed is an unprecedented infrastructure, both domestically and internationally, that provides a trust layer for linkage between AI agents. It consists of a multi-layered defense structure centered around two core components: an 'Agent Store' and a 'Broker Agent'.