#Desilo #Cornami #Encrypted AI

DESILO and Cornami's Breakthrough in Encrypted AI Technology

The world of artificial intelligence (AI) is continuously evolving, and a significant leap forward has come from the collaboration between DESILO, a privacy-enhancing technology startup, and Cornami, noted for its scalable computing architectures. At the AI Infra Summit 2025 in San Jose, California, the two companies proudly announced the deployment of a large language model (LLM) built on Fully Homomorphic Encryption (FHE). This novel technology enables the processing of sensitive information while it remains encrypted, offering both remarkable speed and accuracy suitable for real-world applications.

Addressing Privacy vs. Performance

The AI landscape has long grappled with the dichotomy between safeguarding sensitive data and maintaining robust AI performance. Enterprises in vital sectors like healthcare and finance have traditionally faced tough choices when it comes to protecting their data without compromising functionality. Conventional encryption methods, while strong on privacy, tend to slow down computational processes, while quicker operations often risk exposing sensitive information.

However, this new collaboration between DESILO and Cornami aims to bridge that critical gap, enabling encrypted AI inference speeds that approach those of standard plaintext processing. According to Dr. Craig Gentry, Chief Scientist of Algorithms at Cornami, known as the pioneer of FHE, the introduction of plaintext ciphertext matrix multiplication (PCMM) is a game-changer in this respect. Dr. Gentry emphasizes that PCMM facilitates more than 90% of LLM computations, allowing complex matrix operations—the backbone of LLM processing—to be executed securely without significant delays.

Transformative Applications Across Industries

The implications of this innovation are profound and far-reaching. In healthcare, for instance, clinical trials generate enormous amounts of imaging and patient data that must remain secure for regulatory compliance. The new AI model enables researchers to analyze this encrypted data, preserving patient confidentiality while ensuring timely insights.

The benefits extend beyond healthcare. Financial institutions, government agencies, and cloud services can leverage this technology to meet stringent data sovereignty and zero-trust regulations. Seungmyung Lee, CEO of DESILO, affirms their commitment to the principle of