The father of the internet left Google. And immediately started tackling another fundamental network problem

At the end of June 2026, Vint Cerf ended his twenty-year tenure at Google, where he held the position of Vice President and "Chief Internet Evangelist." But at 83, he is certainly not heading into retirement. As confirmed to TechCrunch, starting July 15 he has been advising the organization Innovation Labs, a subsidiary of the domain registry Identity Digital based in Bellevue, Washington.

"The internet has always evolved by addressing fundamental infrastructure challenges before they become systemic obstacles," Cerf told Forbes. "AI agents are driving the next major architectural evolution, and the questions of identity, accountability, and interoperability require the same thoughtful approach that enabled the internet to scale."

Cerf's name gives the project considerable weight — he co-designed the TCP/IP protocols on which the entire modern internet runs and was present at the creation of ICANN, the organization that oversees the global domain name system. On the DNSid advisory board, he is joined by figures from the fields of internet infrastructure, cybersecurity, national security, and finance.

DNSid: a digital "license plate" for AI agents

The principle of the DNSid project is deliberately understated. Instead of building an entirely new trust system, it builds on technology over forty years old — the Domain Name System (DNS). That's the part of the internet that lets you type a human-readable web address instead of having to know a server's numerical IP address.

DNSid will tie the identity of every AI agent to an existing internet domain and use cryptographic proofs to record how an agent's registration evolves over time. Any organization in the world will then be able to verify who is behind a particular agent — without needing a direct relationship with its operator. This is exactly how DNS works today: independent systems recognize the same identifier without having to know each other.

In June 2026, Innovation Labs submitted a standards proposal to the IETF (Internet Engineering Task Force), which has overseen key internet protocols for decades. According to Innovation Labs' interim CEO Allie Kline, the company is already testing the standard with several unnamed hyperscalers and digital identity companies.

Identity is not trust. It's a license plate, not a background check

It's important to understand what DNSid does — and what it doesn't. The system does not decide which agents to trust or which are authorized to act. It merely identifies who is responsible for an agent. The decision about trust remains with companies, security teams, regulators, and courts.

Forbes captured it with an apt analogy: it's a license plate, not a criminal record extract. When an agent overwhelms someone else's system with requests or performs an unauthorized action, the injured party today often has no way of finding out who is behind it. With DNSid, they would have an "accountability anchor" — they could identify the agent, isolate it, revoke its access, and investigate what happened.

And what about criminals who simply don't register? According to Kline, that's actually an advantage: "The value of a widely adopted standard is that systems can recognize its absence." The inability to verify accountability for an agent should, over time, become a clear risk signal — similar to how a website without HTTPS is viewed today.

Agents are already ordering pizza and managing e-shops

The timing is no coincidence. AI agents are rapidly moving from research demos into the role of real economic actors. On the same day Cerf's role was announced, DoorDash launched a limited beta of its dd-cli tool — a command-line interface through which developers and AI agents can browse stores, fill carts, and order actual food. Adobe, meanwhile, is preparing agentic shopping features for the Adobe Commerce platform used by brands like Coca-Cola.

But according to Forbes, tens of millions of autonomous agents are already moving around the web today with their only "identity" being an IP address. At a time when the number of cyberattacks conducted through AI tools is growing, Cerf sees this as unsustainable. He himself remains a realist, however: "I don't think the agent economy is inevitable. What is inevitable is that people will try it. We are fundamentally lazy creatures — when we find a way for an agent to do something for us, we're very likely to use it."

What this means for Europe and the Czech Republic

For European companies, the topic of AI agent accountability is particularly timely. The EU AI Act already requires that users know they are communicating with artificial intelligence, and for high-risk systems it mandates traceability and human oversight. An open standard that makes it possible to unambiguously identify an agent's operator would significantly ease companies' regulatory compliance — instead of building their own registries, they could simply connect to the global infrastructure.

The domestic context is also interesting: the Czech national domain administrator CZ.NIC has long been a pioneer in DNS security (DNSSEC technology, ODVR resolver). If DNSid gains traction as an open IETF standard, national registries — including the Czech one — could adopt it relatively easily. The key question, according to Cerf, is whether major players will adopt the standard: "No one can do everything you'd want from every agent. We're going to have to rely on user pressure. That's exactly how TCP/IP caught on."

Kline adds a crucial argument: no single company should define how accountability works for everyone else. The open standards that scaled the internet are, in her view, the right template for scaling artificial intelligence as well.