Vint Cerf, one of the architects of TCP/IP and the modern internet, is advising Innovation Labs on a proposal called DNSid, a domain-based registry designed to give AI agents verifiable identities on the open web. Cerf left Google last week after 20 years at the company, and the Innovation Labs role is his first public post-Google engagement. The pitch is straightforward: as autonomous agents begin transacting with each other across company boundaries, the internet needs a way to know which agent is which, who it works for, and who is accountable when it misbehaves.
Innovation Labs is a subsidiary of Identity Digital, a DNS registry company that already operates a large slice of the domain-name system. DNSid would link each agent to an existing internet domain and use cryptographic proofs to log its registration history over time. Interim CEO Allie Kline says the company is trialing the standard with several unnamed hyperscalers and identity firms.
“I felt like I might be able to help them in a period of time when naming and identification is becoming increasingly important.”— Vint Cerf, TCP/IP co-designer and Innovation Labs advisor
The problem DNSid targets is real and getting more urgent. Most AI agents today live inside proprietary systems, calling internal tools for narrow purposes. But vendors from OpenAI to Anthropic to Meta are pushing toward agents that book travel, execute purchases, and negotiate with other agents across organizational lines. Without a shared identity layer, an agent from one platform has no verifiable way to prove to another platform's agent that it is who it claims to be, or that it has the authority to act.
Key facts
- 01Vint Cerf, co-designer of TCP/IP, left Google last week after 20 years at the company.
- 02He is now advising Innovation Labs, a subsidiary of DNS registry firm Identity Digital.
- 03Innovation Labs has proposed DNSid, a registry that ties each AI agent to a domain name with cryptographic proofs.
- 04The group says it is trialing the standard with several unnamed hyperscalers and identity companies.
- 05Innovation Labs pitches DNSid as non-proprietary, arguing hyperscaler-owned standards face 'organ rejection' from the market.
Cerf frames the accountability question as the core of the problem. Agents are far more active than static domains, and it is not yet clear what commitment an organization takes on when it registers one. If an agent placed under a company's identity goes off script and moves money, executes a contract, or leaks data, the identity layer determines who the counterparty can point to.
Kline's argument is that a domain registry is the neutral ground for this. Identity Digital's pitch is that it does not run a foundation model, does not sell agent tooling, and does not want to own the registration data itself. That posture matters because the alternative standards emerging from the frontier labs come with obvious conflicts: a hyperscaler-owned identity system is also a hyperscaler-owned choke point.
Multiple competing proposals are already in flight from vendors trying to define the agent-to-agent protocol layer. Cerf argues the winner will be decided by functionality and user pressure, not by whichever company ships first. He points to the original TCP/IP adoption fight as the template: the protocol won because it was the one that actually let heterogeneous networks talk to each other, and users refused to accept walled gardens that could not interoperate.
That framing puts DNSid in direct competition with any identity scheme that ships from OpenAI, Anthropic, Google, or Meta as part of their own agent stacks. If Company X's agent runs on one vendor's identity system and Company A's agent runs on another, and the two cannot verify each other, the agentic economy stalls at the boundary of each platform. Cerf's bet is that the pressure to interoperate will eventually force convergence on an open standard rather than a proprietary one.
Cerf is careful not to declare the agentic future inevitable, but he is direct about the demand side.
The counterweight to all of this is that identity is the easy part of the agent problem, not the hard part. Even with a verified registry, the harder questions — what authority an agent has been delegated, whether its actions bind the registering organization legally, how to revoke or claw back a rogue agent's transactions — remain unresolved and largely untested in court. DNSid solves naming; it does not solve liability. And competing standards from well-capitalized labs may simply move faster, regardless of how open the alternative is.
Still, the agent-identity fight is one of the most consequential infrastructure debates in AI right now, and it has been running largely without the involvement of the people who built the last identity layer that mattered. Cerf lending his name to a non-hyperscaler proposal shifts the credibility balance, and it gives enterprises weighing whether to wait for an open standard a reason to actually wait. The commercial question is whether Identity Digital can convert that credibility into adoption before the frontier labs ship their own systems and lock customers in by default.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
