Disilo, US Konami Unveil 80x FHE AI Computing Performance Breakthrough

A future where complex AI computations can be performed without exposing sensitive data—this is the world envisioned by 'encrypted AI' based on Fully Homomorphic Encryption (FHE). While hailed as a key technology that simultaneously satisfies privacy and data utilization, the extreme reduction in computational speed in the encrypted state has been the biggest hurdle to its commercialization. However, this significant limitation has finally been overcome. Disilo, a company specializing in privacy-enhancing technologies, has partnered with Konami, a North American computing architecture company, to develop groundbreaking technology to solve this problem. The two companies conducted joint research to redesign the 'encrypted matrix multiplication structure', a key constraint in Fully Homomorphic Encryption environments, and as a result, succeeded in achieving computational performance up to 80 times faster than existing methods. This is the result of the synergy between Disilo's optimization software technology and Konami's scalable high-performance computing architecture. The achievement of this research was recognized for its technical value upon its publication in the ePrint archive of the International Association for Cryptologic Research (IACR), through a paper co-authored by Craig Gentry, the creator of Fully Homomorphic Encryption, and Disilo researcher Lee Yong-woo. Lee Seung-myung, CEO of Disilo, emphasized that this achievement "will be a major turning point for the financial and medical sectors, where AI adoption has been delayed due to concerns about personal data leakage, despite its significant potential benefits," and that it would accelerate the actual deployment of services. Building on this technology development, Disilo plans to accelerate the commercialization of encrypted AI technology, including the establishment of infrastructure enabling secure collaborative analysis between institutions without data movement. This will lay the foundation for a new era where the full potential of AI can be utilized while upholding privacy.