Confidential Computing Summit Pushes 1-Year Attestation Standards for Agentic AI Security
Updated
Updated · InfoWorld · Jun 25
Confidential Computing Summit Pushes 1-Year Attestation Standards for Agentic AI Security
3 articles · Updated · InfoWorld · Jun 25
Summary
San Francisco talks this week centered on agentic AI as confidential computing’s breakout use case, with industry groups framing hardware-backed attestation as the missing trust layer for autonomous agents.
Intel, Microsoft and NVIDIA presented a composite attestation format spanning confidential VMs, CPUs and GPUs, and Intel said the work could advance toward an RFC within 1 year.
Microsoft said the harder task is linking fragmented identity systems to hardware proof, while the industry still lacks standards for agent delegation chains, Model Context Protocol attestation and cross-cloud trust.
Speakers also stressed limits: NVIDIA cited the 2026 TDXRay cache-side technique, Microsoft noted attestation cannot prove data location, and researchers warned hardware flaws are harder to patch.
The push extends a broader Linux Foundation effort to build internet-like trust infrastructure for AI agents, following its recent DNS-based Agent Name Service announcement.
As competing AI identity standards like ANS and AGNTCY emerge, is a new 'standards war' for the agentic web inevitable?
Can the internet's 40-year-old address book, DNS, secure the coming trillion-dollar economy of autonomous AI agents?
The Agent Name Service (ANS): A Scalable Solution for Trusted AI Agent Identity and Governance in 2026
Overview
As autonomous AI agents rapidly proliferate, organizations face major challenges around identity, trust, and governance. The rise of 'shadow AI'—agents without clear identification—creates serious risks, making authentication and robust governance nearly impossible. This lack of transparency has turned identity management into a top security concern. In response, the Linux Foundation introduced the Agent Name Service (ANS), an open standard that builds on the proven Domain Name System (DNS) to provide scalable, verifiable identities for AI agents. ANS aims to address these urgent needs, enabling organizations to securely identify and manage their AI agents in a transparent and trusted way.