DESILO和Craig Gentry通过一项新的FHE计划,大大改进了加密矩阵乘法,从而能够安全、大规模地开展AI行动,在私人人工智能方面迈出了一大步。 DESILO and Craig Gentry unveiled a major leap in private AI with a new FHE scheme that drastically improves encrypted matrix multiplication, enabling secure, large-scale AI operations.

DESILO与FHE发明家Craig Gentry合作,在FHE.org 2026年会议上启动了第五代GL FHE计划,通过大幅提高矩阵乘法效率(AI系统的核心任务),在私人AI方面迈出了一大步。 DESILO, in partnership with FHE inventor Craig Gentry, has launched the fifth-generation GL FHE scheme at the FHE.org 2026 Conference, introducing a major leap in private AI by significantly improving the efficiency of matrix multiplication—the core task in AI systems. 新办法减少了计算间接费用,使人工智能在加密数据上进行安全操作而不破解,这增强了保健、金融和政府等敏感行业的隐私。 The new scheme reduces computational overhead, enabling secure AI operations on encrypted data without decryption, which enhances privacy for sensitive industries like healthcare, finance, and government. 这一进展使实际的、大规模加密的人工智能更接近于现实世界的使用,并在FHE.org上公布了全部技术细节。 This advancement brings practical, large-scale encrypted AI closer to real-world use, with full technical details published on FHE.org.