This can make them a fantastic match for low-belief, multi-bash collaboration scenarios. See here for just a sample demonstrating confidential inferencing determined by unmodified NVIDIA Triton inferencing server. https://one.aistudio.ml