AI Agents Getting Verifiable Identity: What It Means for Your Enterprise

First I argued the economics have to be engineered in. Then I argued identity is the control that makes it accountable. This week the industry caught up to that second point. And not quietly.

This week the internet started building an identity layer for AI agents.

The Linux Foundation announced the Agent Name Service, or ANS. It is a proposed open standard. It gives every agent a verifiable identity, anchored to DNS.

DNS has told your browser which server to trust for thirty years. Now it would tell one agent whether to trust another.

This matters more than a typical standards announcement. Just not for the reason most coverage will give.

What ANS actually proposes

The pitch is simple. No new lookup network. No proprietary registry.

ANS anchors agent identity straight to DNS, the system that already handles more than a hundred million queries a second. You publish your agents through domains you already own.

Any system can then check three things before it interacts. Who the agent represents. What it is allowed to do. Whether its code and history are unchanged.

It is federated. There is no central gatekeeper. It also folds in decentralized identifiers and Legal Entity Identifiers, so existing identity systems map in rather than compete.

Cloudflare, Cisco, Salesforce, and GoDaddy are already behind it. As reported by CIO Dive, it is still at the intent-to-launch stage and is openly seeking input.

Taken at face value, it is a serious answer to a real problem. I have watched plenty of standards arrive late. This one is early, in the right way.

The catch nobody will put on the slide

A standard is not a deployment. ANS is a framework seeking input, not a product you switch on next quarter. Here is why. ANS solves the problem between organizations. It proves an agent’s identity as it moves across the open internet. It does not solve the problem inside your own walls. That is the one your next audit will find.

Picture your underwriting agent acting on a loan file. The question is not whether an outsider can verify it. The question is simpler. Do your own logs show the agent, or a service account it borrowed? That is on you. And it is true today, standard or no standard.

What we do, and what I would do on Monday

On our auto loan and credit card origination work, we settle this on day one. Every agent gets its own identity. Scoped to the steps it owns. Logged under its own name. Nothing inherited. Nothing shared.

That is what makes the system explainable to an examiner.

It is also what makes an agent ready for ANS later, without a rebuild. A clean internal identity can be published to an external trust layer when the time comes. A borrowed service account cannot. You will have to rebuild it first.

So treat this week’s news as a forcing function, not a finish line.

Inventory your agents. Ask which ones have a real identity and which ones borrow. Give the ones that matter their own. Scope them tight. Log them by name. Do that now, and ANS becomes an upgrade you adopt on your terms. Skip it, and the standard arrives to find nothing clean to register.