Revolutionizing Data Security: DESILO Unveils the First Fully Homomorphic Encryption Library for Private AI

Breaking New Ground in Data Privacy with Fully Homomorphic Encryption

In an era where data security is paramount, DESILO has made a significant stride by introducing the first-ever Fully Homomorphic Encryption (FHE) library that incorporates the innovative 5th-generation Gentry-Lee Scheme (GL). This new technology is set to revolutionize the deployment of artificial intelligence (AI) in a secure and privacy-conscious manner, paving the way for what is termed 'Private AI'.

The announcement took place during the FHE.org 2026 Conference held in Taipei, where the creator of FHE, Craig Gentry, along with DESILO's Chief Scientist Yongwoo Lee, co-authored the GL scheme. This breakthrough technology directly addresses the computational challenges faced by earlier iterations of FHE, particularly in executing matrix multiplications—an essential operation in deep learning algorithms.

What Makes the GL Scheme a Game-Changer?

Historically, Fully Homomorphic Encryption has been lauded as the 'holy grail' of data security due to its capacity to allow calculations on encrypted data without needing to decrypt it. However, achieving practical implementations has been cumbersome due to incredibly high computational costs. The GL scheme's architecture optimizes homomorphic operations, particularly for matrix multiplications, thereby significantly reducing computational overhead. This marks a paradigm shift in how AI workloads can be processed while maintaining stringent encryption protocols.

DESILO's new FHE library is not just a theoretical framework; it represents the first commercially viable platform that translates the GL scheme from abstract concepts into high-performance applications. This library, crafted using C++ and CUDA technology, has been optimized for both CPU and NVIDIA GPU deployments, ensuring it can handle the heavy lifting required for modern AI applications. The library also features an innovative dual-scheme setup, incorporating the RNS-CKKS scheme for vector operations and the cutting-edge GL scheme for matrix operations. Moreover, it comes equipped with a Python wrapper, making it incredibly user-friendly for data scientists seeking to implement the library into existing machine learning workflows.

A Closer Look at the Impact on Industries

In sectors where data sensitivity and regulatory compliance are critical—such as finance, healthcare, and enterprise data analytics—this advancement allows for the secure use of AI technologies without compromising client privacy. The release of this FHE library enables organizations to leverage the full disruptive potential of AI without the inherent risks associated with handling sensitive data. Yongwoo Lee emphasizes that, "Matrix multiplication is the dominant workload in modern AI systems. With our new library natively supporting the GL scheme, we are fundamentally restructuring how these critical operations are executed under homomorphic encryption. This innovation bridges the gap between theoretical security and practical AI deployment."

By eliminating the performance barriers associated with encrypted data processing, DESILO is empowering organizations across highly regulated industries to harness AI capabilities like never before. This forward-thinking approach promises to ignite a wave of innovation, as companies increasingly invest in secure data methodologies that comply with global regulations while still promoting operational efficiency.

In conclusion, DESILO's launch of the world's first fully homomorphic encryption library marks a pivotal moment in the evolution of data privacy. As the landscape of data utilization continues to evolve, this technology positions organizations to navigate the delicate balance between innovation and privacy with unprecedented capability. For more information on the DESILO FHE library, follow the links to their technical resources and further insights into the future of Private AI.

To explore more about DESILO and how it is shaping the Private AI landscape, check here.