Keynote: Trustworthy AI with Confidential Computing
summary
Claims an end-to-end confidential AI stack with concrete tools: VEIL implements fully homomorphic encryption for encrypted inference and ships as an SDK used for skin cancer detection, reducing query leakage while remaining quantum safe. PetalGuard combines federated learning with multi‑party computation for decentralized aggregation, claiming zero data exposure, no accuracy loss, and deployments on AWS Marketplace with post‑quantum communications. Prism me provides agent identity and attestation, tying into an Agents Guard control plane for policy and autonomy budgets.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:24Confidential AI necessity
- 2:49Confidential AI pillars
- 5:55Inference with fully homomorphic encryption
- 8:24Private federated learning
- 11:55Agent IAM with PrismMe
- 13:40Future of trustworthy AI
- 14:23Summary
related talks
47mTowards Confidential AI for the Masses!
13mProvably Secure AI for Health Data – Powered by Confidential Computing
36mSecurity for AI and Multi-Party Collaboration with Confidential Computing (Oh, and...
19mHow Can We Have Private Access to Intelligence?
06mSponsored Keynote: How to Protect AI Workloads in Cloud Native
19mPrivate AI: Building Trust Through Verifiable Computation
36mCTO Panel: The Impact of Confidential Computing with Intel, Azure, AMD & NVIDIA
30mKeynote Panel: The Future of...