Source · Tier C · Blog / articleSource · Confidential Inference via Trusted Virtual Machines
Confidential Inference via Trusted Virtual Machines
Anthropic, Pattern Labs. 2025. Anthropic research.
| Original | https://www.anthropic.com/research/confidential-inference-trusted-vms |
|---|---|
| Accessed | 2026-09-25 |
| Note | Published in collaboration with Pattern Labs, with a companion report ("Confidential Inference Systems: Design principles and security risks", June 2025). Sketches a small, attested model loader and invoker that decrypts data only inside a trusted environment and passes it to the accelerator, for protecting model weights and user data; states the work is early. Developer's description of its own research; supports "Anthropic reports" statements. |