TechCrunch: Vint Cerf wants AI agents to have internet identities

Vint Cerf, co-architect of TCP/IP, has joined Innovation Labs to advise on DNSid, a system for assigning verifiable identities to AI agents operating across the internet. The work addresses a gap that becomes more visible as autonomous agents start interacting with external systems on behalf of organizations: when an AI agent sends a request, places an order, or accesses a service, there is no standard way for the receiving system to verify who built the agent, who authorized it, or what it is permitted to do.

DNSid would link agent identities to internet domain names through cryptographic verification, making agents traceable back to their originating organizations. Cerf also identified interoperability as a second gap that the initiative aims to close: agents built by different companies currently do not communicate with each other, which limits the potential for multi-agent workflows that span organizational boundaries.

For product managers building AI products, this work points toward compliance requirements that do not yet exist but are likely coming. Agents operating on the public internet may eventually need to implement identification standards to function reliably across third-party systems. More immediately, trust and accountability are becoming product features in their own right. Enterprise buyers are increasingly asking not just what an agent can do, but who stands behind it, what it can access, and how to audit its actions. Designing for these questions now, before standards are finalized, is easier than retrofitting accountability after a product has launched.