Artificial Intelligence: Vint Cerf Wants to Give Each AI Agent an Identity Card

Vint Cerf, one of the architects of the foundational protocols of the internet (TCP/IP, the infrastructure that allows computers around the world to communicate with each other), is leaving Google after twenty years to advise Innovation Labs. The organization is working on a project called DNSid, which aims to address a question that is becoming central as autonomous AI agents multiply: who is responsible for the actions of a machine that operates without direct human supervision? Key points of this article:

• Vint Cerf, an internet pioneer, has left Google to join the DNSid project aimed at giving a verifiable identity to autonomous AI agents.
• DNSid uses the domain name system to create a "responsibility anchor," allowing for the tracing of the origin and accountability of AI agents' actions.

DNSid, a Passport for Machines

The principle can be summed up in a simple comparison. DNSid relies on the existing domain name system (DNS, the directory that translates a web address into a technical location) to create what the project calls a "responsibility anchor." Specifically, each AI agent is linked to a verified domain name, with cryptographic proof indicating where it comes from, who authorized it, and who is accountable for its actions. A kind of digital passport, except the traveler is not human. Cerf explains his approach candidly: "I thought I might be able to help them at a time when naming and identification are becoming increasingly important," he told TechCrunch on July 15.

The initial observation is not up for debate: AI agents are becoming full participants in the internet, capable of acting through organizations, clouds, governments, and critical infrastructures. Without a durable identity layer, it is impossible to know who is truly behind an automated action. Forbes summarized the issue well on July 17: without a reliable identifier, the question of trust between agents becomes a structural problem rather than just a technical detail.

A problem that crypto is already trying to solve in its own way

This quest for identity is not isolated, and the crypto ecosystem has been tackling it in parallel for months. Le Journal du Coin detailed the new ERC-8004 standard on Ethereum, which also relies on three registers: identity, reputation, and validation, to allow agents to prove their reliability as they move from one network to another. The same logic applies to Ledger, which has developed a hardware identity for each agent, recorded directly on a blockchain. On one side, DNSid leverages the historical infrastructure of the web. On the other, crypto relies on decentralized registers. Two approaches, one shared diagnosis: without a verifiable identity, an AI agent remains a black box that cannot be audited or held accountable.

However, DNSid starts with an advantage that crypto has not yet achieved: DNS is already installed everywhere, on every connected device, without a user needing to understand anything about blockchain to benefit from it. If the project delivers on its promises, it could become the default authentication layer of a web populated by autonomous agents, leaving on-chain registers to play a more specialized role, reserved for financial uses and transactions that require immutable proof rather than just a name verification.