SEOUL, South Korea, Sept. 2, 2025 /PRNewswire/ — Desilo Inc., a privacy-technology startup specializing in cryptographic data collaboration, announced plans to launch its new healthcare platform, HARVEST™, this December. The platform is designed to allow medical institutions and pharmaceutical companies to collaborate on research without exposing sensitive patient data.
HARVEST™ by Desilo lets healthcare institutions collaborate on analysis and AI training—without ever exposing raw patient data.
Medical research often requires large, multi-institutional datasets, but privacy regulations and security risks have made data sharing difficult. According to IBM, the average cost of a healthcare data breach is $10.93 million, while violations of Europe’s GDPR can trigger penalties of up to €20 million. Desilo says HARVEST™ addresses this challenge by applying Fully Homomorphic Encryption (FHE) and Federated Learning (FL) to enable joint analysis and model training while keeping raw data encrypted at all times.
The company says studies that previously took months could be completed in about a week using HARVEST™, significantly improving research efficiency. The platform supports horizontal and vertical FL, split learning, and cohort federation, while automatically managing compliance requirements under GDPR and HIPAA.
"Pharmaceutical companies and research consortia often face a tradeoff between privacy and progress," said Seungmyung Lee, Chief Executive of Desilo. "Our mission is to shorten research timelines and reduce development costs—ultimately helping accelerate drug development and diagnoses without compromising patient privacy."
Desilo’s technology has already drawn attention. The company won first place at iDASH, a leading global competition in privacy-preserving genomics, and has developed its own production-grade FHE library. In collaboration with U.S.-based hardware innovator Cornami, Desilo also built what it calls the world’s first commercial-scale encrypted large language model, leveraging GPU acceleration to address performance barriers in homomorphic encryption.
The startup has also been selected for several government-backed initiatives, including Korea’s National Cancer Center Safe Data Zone Project.
Desilo plans to make HARVEST™ generally available in December, with early demonstration access available through contact@desilo.ai.