Cornami and DESILO Unveil Scalable Encrypted AI with Revolutionary FHE-Based LLM

Cornami and DESILO Unveil Scalable Encrypted AI

At the recent AI Infra Summit 2025 in San Jose, California, two pioneering companies in the tech industry, Cornami and DESILO, revealed a significant advancement in encrypted artificial intelligence (AI). The collaboration resulted in a deployable large language model (LLM) that utilizes fully homomorphic encryption (FHE) technology. This innovative approach enables the processing of sensitive data while maintaining encryption, thus achieving rapid real-world performance and accuracy that could redefine how enterprises manage and protect their data.

Breaking the Privacy vs. Performance Barrier

For decades, organizations across various sectors have grappled with the challenge of balancing data privacy and performance. Enhanced security measures often lead to slower AI system responses, while a focus on speed detracts from data protection. However, the partnership between Cornami and DESILO seeks to resolve this long-standing conflict. Their new solution allows encrypted AI inference to occur at speeds nearly equivalent to unencrypted operations.

Dr. Craig Gentry, the Chief Scientist of Algorithms at Cornami and a pivotal figure in the development of FHE, explained, "For years, FHE was deemed too slow for practical applications. Our breakthrough in accelerating encrypted computations demonstrates that businesses can achieve both privacy and performance without compromise." One innovation that significantly contributes to this capability is the plaintext ciphertext matrix multiplication (PCMM), which impacts over 90% of LLM computations. By enabling secure execution of matrix multiplication with minimal overhead, PCMM optimizes the performance of AI systems utilizing encryption.

This technological advancement, combined with Cornami’s high-performance Compute Fabric, results in a dramatic improvement in efficiency, exceeding traditional encrypted computation methods by several orders of magnitude. Dr. Gentry added, "The value of PCMM is in its ability to make privacy-preserving AI feasible, ensuring compliance and post-quantum security while closing the gap in performance between encrypted and plaintext AI inference."

Transforming Industries with Encrypted AI

The implications of this technology are vast, particularly within the healthcare sector. Clinical trials often generate massive amounts of sensitive patient and imaging data that require robust security. This new development allows researchers and clinicians to analyze encrypted data without compromising patient confidentiality or data integrity, enabling timely insights that are crucial in healthcare contexts.

Besides healthcare, sectors such as finance, government, and cloud services stand to benefit immensely. With increasing regulatory pressures demanding stricter compliance, organizations can now leverage encrypted AI without sacrificing the speed of their operations. As Seungmyung Lee, CEO of DESILO, pointed out, "Our guiding principle has always been to simplify the process by avoiding decryption. Our collaboration with Cornami makes that ideal a realizable goal. This partnership sets the stage for enterprises to safely access valuable data and aligns with the upcoming launch of our HARVEST platform, which will cater to global healthcare partners."

About Cornami and DESILO

Cornami specializes in high-performance, scalable computing solutions that allow advanced encryption technologies, like FHE, to function at unprecedented speeds. Their architecture is built to support privacy-preserving AI across industries demanding both security and efficiency. On the other hand, DESILO, a South Korean startup focused on privacy-enhancing technology, is expanding its footprint globally. It offers secure collaborative solutions leveraging FHE and federated learning, ensuring that sensitive data remains encrypted throughout the analysis process. Their flagship platform, HARVEST, is anticipated to launch in December 2025, further advancing secure multi-party data collaboration in the healthcare sector and beyond.

With the advancements introduced by Cornami and DESILO, the landscape of AI technology is set to shift, potentially ushering in a new era where privacy concerns are elegantly balanced with unprecedented performance enhancement.