Confidential AI: Enabling Trustworthy Ecosystems in the Age of Agentic Intelligence - Session
summary
Argument: use hardware TEEs to protect data-in-use, not just rest/transit; claim is that Intel SCX/TDX plus attestation and TDX Connect for GPUs let enterprises run GenAI and federated training without exposing prompts or memory contents. Demo shows a plain memory dump revealing prompts then an Intel TDX deployment where the memory dump hides prompts; project integrations include OpenFL and Open Platform for Enterprise AI and claim ~5% overhead versus 20–30% for pure crypto approaches.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 2:12Why confidential AI is essential
- 6:02Confidential computing foundation
- 8:10Agentic AI vulnerabilities
- 11:40What confidential computing is
- 15:50Intel’s confidential AI solutions
- 18:16Healthcare and autonomous driving spotlight
- 21:24Open Platform for Enterprise AI demo
- 22:29Live demo: memory dump risks
- 23:10Confidential deployment for memory
- 23:52Performance impact and trust
- 25:35Trust via attestation and compliance
- 27:41Industry collaboration and future
- 28:22Q&A
related talks
46mOpen for AI: Secure paths to data collaboration volume and diversity
42mSponsored Session: Confidential Computing - New Capabilities for New Workloads- 42mSponsored Session: Confidential Computing - New Capabilities for New Workloads
28mGenerative AI + Proprietary Data
19mHow Can We Have Private Access to Intelligence?
37mIt's All About Trust: Why Confidential Computing Matters
20mKeynote: Future-Proofing Confidential AI
24mStrictly Confidential: Use Cases & Demos of AI Workflows in TEEs