Linux Foundation Launches Agent Name Service to Give AI Agents a Verifiable Identity

How do you know the AI agent asking for your payment details is really your agent? That question is getting harder to dodge. Agents now book flights, buy software, message suppliers and move through company systems, often with nobody watching. On November 10, 2026, the Linux Foundation welcomed a project built to answer it: the Agent Name Service, or ANS, an open standard that gives AI agents a verifiable identity, a way to be discovered, and a way to authenticate each other.
Why agents need an identity you can check
Think about what an agent does on your behalf. It might reserve a hotel room, pay an invoice, reorder stock or negotiate with another company’s agent at three in the morning. Every one of those actions runs on trust. If a stranger’s agent can simply claim to be yours, the whole arrangement falls apart fast.
We’ve solved a version of this problem before. DNS gave websites names that machines could resolve and people could remember. Certificates then let your browser check that a site really is who it claims to be. ANS takes a similar route for agents: anchor the identity to a domain, so an agent’s name traces back to an organisation that controls that domain and can be held responsible for it.
The timing isn’t accidental. Airlines have already moved to block unauthorised AI agents from booking flights, and security researchers keep finding fresh ways for agents to be manipulated, including backdoors hidden inside agent systems. The industry is deploying agents first and asking identity questions later. ANS is an attempt to reverse that order.
How the Agent Name Service works
ANS is built on existing DNS infrastructure instead of a brand-new network. As described in the announcement, it covers domain-anchored naming, registration, DNS publication, public-log auditing, integrity monitoring and agent-to-agent authentication.
In plain terms, an agent gets a name tied to a domain it belongs to. That name gets published through DNS, the same plumbing that already directs much of the internet’s traffic. Other software can then look the agent up before trusting it: is this name registered, does the record match what was published, has anything changed quietly? The public log matters here, because a registration that changes leaves a trail anyone can audit.
Discovery is part of the design too. An agent that can be found and verified through a standard lookup is far easier to work with than one introduced by a random API key pasted into an email. And when two agents meet, authentication runs both ways, so each side can check the other before any data or money moves.
Does a verified identity fix every agent problem? No. An agent can be perfectly genuine and still make a terrible decision. But impersonation is the hole attackers are walking through right now, and a checkable identity makes that hole much smaller.
Who is backing the project
ANS arrives with seven specification documents and support from 11 organisations, including Cisco, Cloudflare, DigiCert, GoDaddy, Salesforce and OWASP. That’s a telling mix. Networking, DNS registration, digital certificates and application security are exactly the trades you’d want in the room for an identity system built on domain infrastructure.
The Linux Foundation announced the project in a November 10, 2026 release distributed through PR Newswire, with coverage from Unite.AI and ITBrief following the same week. Placing the project at the Linux Foundation signals intent, too: this is meant to be an open standard anyone can implement, not a single vendor’s login system wearing a new hat.
The honest catch: the specs are still drafts
Here’s where I pump the brakes. Those seven specification documents are still labelled drafts. Details can change, implementations are young, and nobody should treat ANS as a finished shield they can switch on today.
Standards also live or die on adoption. DNS worked because everyone used it. ANS only matters if agent platforms, cloud providers and ordinary companies actually register their agents and check identities at the door. Big-name backers give it a real chance, but a draft with famous logos is still a draft.
What to watch next
The signals worth watching are practical ones. Do major agent platforms start publishing ANS records for their agents? Do registries and certificate providers build registration into tools people already use? And do businesses start refusing connections from agents they can’t verify, the way browsers now warn about sites with broken certificates?
If even part of that happens, “who is this agent, really?” stops being a guessing game. For a technology that’s increasingly allowed to spend money and speak in your name, that’s not a small thing.
