cd /news/ai-agents/ai-agents-have-an-identity-problem-t… · home › topics › ai-agents › article
[ARTICLE · art-172698] src=byteiota.com ↗ pub= topic=ai-agents verified=true sentiment=↑ positive

AI Agents Have an Identity Problem. The Linux Foundation’s Answer Is DNS.

The Linux Foundation introduced the Agent Name Service (ANS) at KubeCon last week, a standard that ties each AI agent's identity to a DNS domain name its operator already owns, using two DNS TXT records and a DNSSEC-signed transparency log. GoDaddy built the original ANS specification and donated it to the Linux Foundation after submitting it to the IETF as draft-narajala-courtney-ansv2; eleven organizations, including Cisco, Cloudflare, DigiCert, GoDaddy, Infoblox, Radix, Salesforce and OWASP, have contributed to or adopted the standard. The move addresses a gap highlighted by a security scan of roughly 2,000 public MCP servers earlier this year that found essentially none enforced authentication, letting any agent that knows an endpoint claim to be a trusted service.

key takeaway

ANS anchors agent identity to two DNSSEC-signed TXT records — a discovery record under _ans.{agent-domain} and a badge record under _ans-badge.{agent-domain} pointing to an append-only Merkle-tree transparency log — so operators can prove an agent's authorized version and revocation state without a

read4 min views4 publishedNov 19, 2026
AI Agents Have an Identity Problem. The Linux Foundation’s Answer Is DNS.
Image: Byteiota (auto-discovered)

Your AI agents can call other agents. What they cannot do is prove who they are talking to. A security scan of roughly 2,000 public MCP servers earlier this year found that essentially none enforced authentication — meaning any agent that knows an endpoint can claim to be a trusted service. The Linux Foundation stepped in at KubeCon last week with Agent Name Service (ANS), a standard that anchors every AI agent’s identity to a DNS domain name the operator already owns. No new namespace. No proprietary registry. No gatekeeper.

DNS as the Trust Anchor #

The core idea is elegant in its restraint: every agent gets a Fully Qualified Domain Name tied to a domain its operator controls. That domain ownership becomes the root of trust. Because you already proved you own example.com when you registered it, ANS uses that existing proof rather than inventing a new credential system from scratch.

In practice, two DNS TXT records do the work. A discovery record under _ans.{agent-domain} tells any client where to find the agent card and what protocol the agent speaks:

_ans.payments-agent.acme.com IN TXT "v=ans1; version=v1.2.0; p=a2a; url=https://payments-agent.acme.com/.well-known/agent-card.json"

A badge record under _ans-badge.{agent-domain} points to the transparency log entry for that agent. DNSSEC signs the entire chain, making the records tamper-evident. The design runs on every DNS provider today with no custom extensions — Phase 1 is intentionally boring infrastructure.

The Transparency Log Is What Makes This Serious #

Publishing DNS records is easy to fake if you have domain access. The ANS transparency log is what separates this from putting a JSON file at /.well-known/agent.json and calling it identity. Every lifecycle event — registration, version update, revocation — gets sealed into an append-only Merkle-tree ledger. Entries get COSE/SCITT receipts: cryptographic proof of an agent’s state at a specific moment in time. You cannot alter or remove those entries after the fact.

That matters for enterprise and regulated industries in a way that “trust the endpoint” never will. When a compliance audit asks whether the billing agent that approved a transaction last Tuesday was actually authorized and running the signed version, ANS makes that question answerable. That is the kind of audit trail that gets enterprise security teams to sign off on deploying autonomous agents.

ANS and DNS-AID Are Not the Same Thing #

The community has been understandably confused by two overlapping DNS-based agent standards. Here is the distinction: ANS answers “can I trust this agent?” DNS-AID answers “where do I connect to this agent?” They are designed to work together, not compete.

ANS DNS-AID
Primary question Can I trust this agent? Where do I connect?
DNS records used TXT ( _ans ,_ans-badge ) SVCB ( _ag )
Primary focus Identity and audit trail Capabilities and endpoints

ANS’s default discovery profile actually uses DNS-AID’s SVCB format under the hood. The DAWN working group at IETF, which handles agent discovery, explicitly excluded identity from its initial scope. ANS fills that gap. Use both if you are building agents that cross organizational boundaries.

Eleven Organizations and a GoDaddy Origin Story #

GoDaddy built the original ANS specification and donated it to the Linux Foundation after taking it to the IETF as draft-narajala-courtney-ansv2. The eleven organizations that have contributed to or adopted the standard — Cisco, Cloudflare, DigiCert, GoDaddy, Infoblox, Radix, Salesforce, OWASP, DistributedApps.ai, HOL, and Trustmodel.ai — include the infrastructure companies that actually run the internet’s naming and certificate infrastructure. When Cloudflare, Infoblox, and DigiCert back a DNS-based identity standard, the chance of it becoming the floor for agent authentication is significantly higher than with most open standards proposals. The ANS registry at GoDaddy is live today for self-service registration.

What ANS Does Not Do — and Why That Matters #

ANS verifies identity, not behavior. It tells you that the agent at payments-agent.acme.com is registered, is running version 1.2.0, and that version was sealed into the transparency log. It does not tell you that the agent will not exfiltrate your data or make unauthorized API calls. Identity is a floor, not a ceiling. Additionally, DNSSEC adoption sits at roughly 30% of domains globally, which means the full cryptographic trust chain will not be available for every agent deployment from day one. Domain hijacking also becomes agent identity hijacking — the security of your registrar account is now part of your agent security posture.

What to Do Now #

If you are building agents that other organizations’ agents will call, registering in ANS is worth doing now — the process uses the same ACME DNS challenge mechanics as Let’s Encrypt. The Go SDK at github.com/agentnameservice/ans-sdk-go handles registration, verification, and transparency log queries. For agents operating only within your own infrastructure, SPIFFE/SVID still makes more sense. ANS is for the boundary: the calls that cross organizational lines, where you cannot assume a shared trust domain. As multi-agent architectures become the default and not the exception, that boundary is about to get a lot of traffic.

── more in #ai-agents 4 stories · sorted by recency
── more on @linux foundation 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/ai-agents-have-an-id…] indexed:0 read:4min 2026-11-19 · —