Linux Foundation launches Agent Name Service to give AI agents verifiable DNS-based identity

The Linux Foundation announced plans to launch the Agent Name Service (ANS), an open standard that grants AI agents verifiable identities through the internet's domain name system (DNS).

The concept builds on research published by the OWASP GenAI Security Project in May 2025, authored by Ken Huang of DistributedApps.ai and Akram Sheriff at Cisco. It has evolved from earlier versions—the original proposed a centralized registry with DNS-inspired naming, while the current draft, submitted to the Internet Engineering Task Force in April, anchors each agent to a domain already controlled by its operator.

How ANS works#

The system mirrors how websites establish identity today. An operator proves control of a domain (say, example.com) using ACME—the automated protocol behind Let's Encrypt—and a registration authority issues the agent a pair of certificates. Every status change (registration, renewal, revocation) is logged in an append-only ledger. Clients verifying an agent can select their assurance level, from basic certificate validation to checking the audit log.

Notably, ANS separates identity from discovery, delegating agent-finding functions to other services.

Discovery and competing frameworks#

The Foundation also adopted DNS-AID, a separate discovery standard developed by Infoblox that lets agents publish endpoints as DNS records for peer discovery. GoDaddy, also involved in ANS, backs DNS-AID.

However, agent identity schemes built on DNS aren't exclusive to these two. At least four similar proposals now exist—including DNSid from registry operator Identity Digital and AID, a minimal community-driven discovery draft. As OWASP's Vineeth Sai Narajala noted, the goal was "extending the foundational trust of the internet to a new generation of autonomous technology," not starting from scratch.

Yet relying on existing DNS registrars and certificate authorities means inheriting their trust hierarchies, which security researchers have long flagged as fragile.

The bigger picture: competing standards#

The Linux Foundation is backing several alternative agent identity systems in parallel. Google's A2A protocol issues signed "Agent Cards" publishable at known URLs, with a registry in development. Cisco's offers its own directory and cryptographic identity layer. Outside the Foundation, Microsoft's Entra Agent ID and —both available since spring—treat agents as corporate directory identities with short-lived, action-bound tokens.