Vint Cerf is working on a plan to unleash AI agents on the open internet
TechCrunch Tim Fernholz
Vint Cerf, one of the internet's original architects, is now advising a group building ID systems for AI agents. The goal: let agents prove who they are online, before autonomous bots start running wild.
Based on reporting by TechCrunch, Tim Fernholz — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Vint Cerf just left Google after two decades there, and instead of taking a breather, he's jumped into one of the internet's next messy problems: how do you know an AI agent is who it says it is? Starting today, he's advising Innovation Labs, a subsidiary of domain registry company Identity Digital, on a proposed standard called DNSid that ties agent identities to existing internet domain names using cryptographic proofs.
Right now, most AI agents operate inside walled gardens, pulling from internal company resources for narrow tasks. But businesses are already picturing something bigger: agents roaming the open internet, talking to other agents, making decisions with real consequences. The problem is nobody has agreed on how to identify or audit those agents once they're out there acting semi-independently.
Cerf isn't alone in this effort. He joins a handful of other internet veterans backing Innovation Labs, whose interim CEO Allie Kline says the company is already testing its approach with several hyperscalers and identity companies she declined to name. Kline argues that Innovation Labs' pitch is that it doesn't want to own the registration data or build out other AI products around it — a deliberate move, she says, because companies tend to distrust a hyperscaler controlling proprietary identity data for the whole ecosystem.
Cerf frames the stakes in terms of trust and accountability: who granted an agent its authority, who answers for what it does, and why anyone should believe its claimed identity in the first place. He compares the moment to the early, chaotic days of networking protocols, betting that whichever standard wins will do so because users demand interoperability, the same dynamic that pushed TCP/IP into dominance decades ago. Multiple competing proposals for agent identity are emerging simultaneously, and Cerf expects the sorting-out process to be genuinely difficult, given how much more autonomous and active agents are compared to a static domain name.
Asked point blank whether an agent-driven internet is inevitable, Cerf pushed back a little. He doesn't think the outcome is guaranteed, but he's confident people will keep pushing toward it anyway, simply because humans gravitate toward whatever's easier, and handing tasks off to an agent is, for most people, easier.
My take — AI-written commentary, not fact-checked reporting
Handing Identity Digital the job of deciding how AI agents prove who they are is a bit like letting the landlord write the building code, except this time at least they're promising not to hoard the data or build a side business on it. That's the right instinct, because nobody wants a single hyperscaler owning the identity layer for every agent on the internet. Cerf's TCP/IP comparison is apt, but worth remembering: that standard won because it was open and dumb-simple, not because a famous person backed it.
Read more about this at: TechCrunch