Vint Cerf backs an identity layer for the agentic internet

AI agents will struggle to operate safely across the open internet unless other systems can verify who controls them and what authority they have. Internet pioneer Vint Cerf is backing an early attempt to build that identity layer.

What happened

Cerf joined Innovation Labs as an adviser on DNSid, a proposed identification standard for autonomous AI agents. The project would connect agents to existing internet domains and use cryptographic proofs to establish registration history and ownership.

The idea is to let websites, services and other agents determine whether an automated system is associated with a legitimate organisation. Innovation Labs says it is testing the approach with unnamed hyperscalers and identity companies.

DNSid is still an early proposal rather than an adopted internet standard. Details about governance, interoperability and broad industry support remain limited.

Why it matters

Agents can already browse websites, call APIs and transact through software. But an open agent economy requires more than technical connectivity. Participants need ways to answer basic questions: Who operates this agent? What is it allowed to do? Can its authority be revoked? Who is accountable if it causes harm?

Using domains could make identity easier to understand because organisations already control and manage them. Cryptographic proofs could add stronger verification than a simple self-declared name.

The bigger picture

Agent identity is becoming a protocol problem. Several companies and standards groups are developing competing approaches to authentication, permissions and reputation.

The winning system may need to be open, interoperable and independent enough that no single model or cloud provider controls it. DNSid’s connection to existing internet infrastructure is strategically interesting, but its credibility will depend on neutral governance and adoption across rival platforms—not only technical design or high-profile advisers.