Confidential Computing on GPUs
NVIDIA H100 Confidential Computing Mode Architecture
NVIDIA's H100 encrypts GPU memory to shield AI model weights and data during computation.
Lorena Sundqvist
· 11 min read
Private AI Review — Page 8
Confidential AI Inference at Scale
Threat Model for a Confidential LLM Inference Endpoint
Confidential computing alone cannot prevent all attacks on encrypted LLM inference.
Kian Vance11 min read
Private AI Use Cases and Configuration
Private AI for Source Protection in Investigative Journalism
Four structural risks in cloud AI systems threaten journalists' source protection.
Kian Vance12 min read
