Cornami and DESILO Break Privacy-Performance Barrier for Encrypted AI
At the AI Infra Summit 2025, Cornami, a leader in scalable computing architectures, teamed up with privacy startup DESILO to launch a fully homomorphic encryption (FHE) based large language model (LLM). The platform can process sensitive data without ever decrypting it, achieving near-plaintext speeds—a breakthrough for enterprises balancing security and performance.
For years, industries from healthcare to finance have faced a persistent tradeoff: stricter data protection slows AI, while faster insights risk exposing sensitive information. This collaboration promises to end that dilemma.
“For decades, FHE was considered too slow for real deployment. By accelerating encrypted computation, we are proving that enterprises can have both privacy and performance,” said Dr. Craig Gentry, Chief Scientist of Algorithms at Cornami and the father of FHE.
Key Innovation: Plaintext Ciphertext Matrix Multiplication (PCMM)
Cornami’s innovation, PCMM, optimizes over 90% of LLM computations—matrix multiplications—allowing them to be performed securely and efficiently. Coupled with Cornami’s Compute Fabric, this delivers orders-of-magnitude faster encrypted AI performance than traditional approaches.
“PCMM makes privacy-preserving AI practical, closing the performance gap between fully encrypted and plaintext LLM inference,” explained Gentry. “This strengthens compliance, sovereignty, and post-quantum security.”
Real-World Applications Across Industries
Healthcare is an immediate beneficiary. Clinical trials generate massive amounts of imaging and patient data that must remain secure, yet timely insights are essential. With FHE-powered LLMs, researchers can analyze encrypted patient records without ever exposing sensitive information.
The same approach extends to finance, government, and cloud services, where zero-trust and data sovereignty requirements are critical. Seungmyung Lee, CEO of DESILO, added:
“Our principle has always been simple: never decrypt. With Cornami, we are making that principle practical. This lays the foundation for enterprises to safely unlock high-value data, aligning with the upcoming HARVEST platform for global healthcare partners.”
Implications for Enterprise AI
This development signals a turning point for privacy-preserving AI, making fully encrypted, high-performance LLMs feasible for real-world use. By bridging the gap between compliance and efficiency, Cornami and DESILO are enabling enterprises to unlock sensitive data for actionable insights without compromising security—a potential game-changer for industries managing high-stakes data.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
