The open internet lacks a standard for AI agents to identify themselves, which blocks them from operating autonomously across different systems. Most agents today stay locked inside proprietary platforms, unable to interact with agents from other companies. Vint Cerf, one of the architects of the internet’s core protocols, has joined Innovation Labs to tackle this problem, warning that without shared identity and accountability, the agent economy will fragment into non-interoperable silos. He draws a direct parallel to the early internet: no single company can build every possible agent capability, so users will ultimately force the industry toward open standards.

Innovation Labs, a subsidiary of Identity Digital, proposes DNSid, a registry that ties each AI agent to an existing internet domain name and uses cryptographic proofs to log its registration over time. The approach leverages domain-name infrastructure as a practical accountability layer — when an agent acts, someone owns the domain behind it. The company is trialing the standard with several unnamed hyperscalers and identity firms, and Cerf believes the key to adoption is functionality, not vendor control. Kline, the interim CEO, explicitly says the proposal avoids locking registration data inside a single hyperscaler, which she calls a source of “organ rejection” for proprietary standards.

Builders should watch whether agent identity becomes a gatekeeping layer or a genuine open utility. Cerf’s involvement signals that this is not a niche infrastructure play but a foundational question for the next era of the internet. The hardest problem is not technical — it’s that no one yet knows what commitment an organization makes when it registers an agent, and agents are far more dynamic than static domain names. For now, the smart move is to plan for multi-agent interoperability from day one, and to keep an eye on whether DNSid or a competing standard gains real deployment outside controlled trials.