DESILO

DESILO develops privacy-enhancing technology (PET) for secure data collaboration and AI applications, with a primary focus on homomorphic encryption (HE). The company builds infrastructure that allows organizations to analyze, share, and utilize sensitive data, including data used in AI model training and inference, without exposing the underlying raw data. Core technical work includes development of homomorphic encryption libraries (Liberate.FHE), a Data Clean Room product, and research into new HE schemes and hardware acceleration, including collaboration with FHE researcher Craig Gentry on a next-generation encryption scheme (GL Scheme) and an end-to-end private LLM inference system (THOR) presented at ACM CCS 2025. DESILO works with financial institutions, universities, and research organizations in South Korea, including Hana Bank, Shinhan, Banksalad, Korea Credit Data, and Seoul National University, as well as technology partners such as Microsoft, NVIDIA, and Cornami. Government-funded R&D projects the company has participated in cover differential privacy for national statistics, homomorphic encryption-based credit scoring, and privacy-preserving cloud AI systems (K-CLOUD). The company's offerings target sectors that handle sensitive personal or financial data, such as banking, healthcare genomics, and data science research, where regulatory and privacy requirements restrict direct data sharing. DESILO's products are positioned as infrastructure for enabling data use across organizational boundaries while maintaining data confidentiality through cryptographic methods rather than access controls or anonymization alone. DESILO was founded in February 2020 and is headquartered in South Korea, with funding from investors including Bon Angels, KB Investment, LG Electronics, NAVER D2SF, and the Schmidt family office.</description> <parameter name="summary">Builds homomorphic-encryption-based Private AI infrastructure for secure data collaboration.