Agentic ID? Vint Cerf Joins Project To Give Every AI Agent A Durable Identifier

Vint Cerf, one of the fathers of the internet, is backing a bold new standard: a durable identifier for every AI agent.

The 'Agentic ID' project, led by the nonprofit TrustoverIP Foundation, aims to give every autonomous AI agent a persistent, verifiable identity — much like domain names or email addresses today. When your agent talks to another agent, the receiving system can immediately check the ID to confirm who (or what) sent the message, and whether that entity is authorized. The initiative comes as AI agents — from personal assistants to enterprise automation bots — proliferate without any universal way to authenticate them.

Why now? The explosion of large language models and agentic AI systems has created a Wild West of autonomous entities interacting with users, websites, and each other. Fraud, spam, and misinformation amplified by rogue agents are growing concerns. Without an identity layer, there's no way to hold an AI agent accountable for its actions, or to build trust between agents.

The project builds on decades of work in internet identity. Vint Cerf, who co-designed TCP/IP and now serves as Google's Chief Internet Evangelist, joins as an advisor. The technical approach mirrors the Domain Name System (DNS): a decentralized registry of agent IDs linked to cryptographic proofs of ownership. Each agent would be issued a unique identifier that persists across sessions, platforms, and jurisdictions — much like a passport for software.

Key figures include the TrustoverIP Foundation's CEO, Heather Flanagan, and a technical steering group from companies like IBM, Microsoft, and the Linux Foundation. The first draft of the standard is expected by early 2027. Early trial deployments are planned in healthcare, finance, and government services where agent-to-agent communication requires audit trails and compliance.

'The problem of agent identity is existential for the AI economy,' says Dr. Jamila Hasan, a researcher at Stanford’s HAI Institute. 'If we can't trust which agent is speaking, we can't trust the information it provides or the transaction it executes.' The initiative touches on deep questions about AI accountability, liability, and even whether agents should have legal 'personhood' of a sort.

Looking ahead, the Agentic ID project could become the foundational layer for AI interoperability — much like TCP/IP became the backbone of the internet. But it faces hurdles: adoption at scale, resistance from companies that benefit from anonymous agents, and the sheer complexity of governing autonomous identities. If successful, every AI agent could eventually be traced, trusted, and responsible.