ทวีตที่ปักหมุด
@NVIDIA and @Protopia_AI are addressing the Trust Dilemma together, unlocking confidential AI anywhere at global scale. protopia.ai/zero-trust-ai-…
NVIDIA Confidential Computing protects the model and its execution environment. Protopia's Stained Glass Transform protects the inference data path across the full operational surface of the AI factory, so sensitive prompts never appear in plaintext in logs, observability, scheduler records, or fabric traffic.
With CC on the weights and SGT on the data path, the operator holds neither party's secrets in the clear.
This fundamentally changes what becomes architecturally possible for sensitive agentic workloads. The secure agent workspace reference design treats inference as a hard dependency that can be served either locally on a nearby GPU or routed to a managed endpoint, for example an AI Factory. Until now, routing sensitive agentic workflows to an LLM in multi-tenant environments have not been fully trusted.
This traditionally results in an isolation tax that eats into the ROI of AI Factories. With the joint stack in place, inference on a multi-tenant AI Factory becomes a viable option for sensitive agent workloads, which widens the set of deployment patterns enterprises can use to put agents into production.
If you're at GTC Taipei this week, come see it live!
GIF
English
