AI

Linux Foundation extends DNS to AI agents with new Agent Name Service

The Linux Foundation said today it plans to launch the Agent Name Service, an open standard that gives artificial intelligence agents trusted identities through the same Domain Name System that already runs the internet.

Using ANS, a system or a person can confirm what organization an agent belongs to and what it has permission to do. It can also show whether the agent’s code or its record of past activity has been altered. There is no new lookup network and no proprietary registry. Identity is tied straight to DNS, which handles more than 100 million queries per second worldwide.

The timing reflects how quickly agents are reaching live systems. Enterprises are putting agents into production before they have settled how to authenticate them, govern what they do or make them work across systems. The Linux Foundation pointed to World Economic Forum data showing 82% of executives plan to adopt AI agents within one to three years. Many of those same executives are not confident they can vet or control the agents once they are running.

Jim Zemlin, chief executive of the Linux Foundation, tied the standard to that gap. Agents are about to operate across companies and platforms, he said and that makes verified identity something organizations have to get right early rather than bolt on later. He argued that anchoring the framework to DNS and open standards makes it possible to scale verified agent communication across the wider digital economy.

ANS supports decentralized identifiers and Legal Entity Identifiers, allowing organizations to fold existing identity systems into a single verification model. The Linux Foundation is positioning the project as vendor-neutral, an argument it has leaned on before in standards work and one that several backers echoed.

The launch arrives with industry support attached. Jared Sine, chief strategy and legal officer at GoDaddy Inc., said building on proven internet foundations gives agents a way to be identified across the open web without recreating the walled gardens of earlier platform shifts. Dane Knecht, chief technology officer at Cloudflare Inc., said extending DNS to agents addresses the identity and security problem “before it gets out of hand.” Cisco Systems Inc. and Salesforce Inc. also signaled involvement, with Cisco saying it is contributing to open standards efforts at both the Internet Engineering Task Force and the Linux Foundation.

The standard grew out of an award-winning research paper led by Ken Huang, chief executive of DistributedApps.ai, with co-author Vineeth Sai Narajala of the Open Worldwide Application Security Project. Huang said his concern had been that agents would proliferate without a neutral identity and discovery layer, creating the shadow AI risks that worry security teams.

The Linux Foundation is asking enterprises, developers, infrastructure providers and security researchers to contribute. Technical repositories and contribution details are available through the Agent Name Service organization on GitHub.

The push extends a busy stretch of work on agent trust. Diagrid Inc. this month shipped cryptographic execution proofs for agents, and startup Tenet Security Inc. launched a platform to catch rogue agent behavior at runtime.

Image: Linux Foundation