Why AI agents need IDs
Cerf's argument centers on the idea that as artificial intelligence becomes more capable of acting independently — booking travel, managing finances, negotiating contracts — those actions need to be verifiable. An AI agent should be able to prove who it is, who authorized it, and what it's allowed to do. That requires a standardized identity layer, much like the domain name system (DNS) that Cerf helped create for the internet.
He has been advocating for a framework that would give each autonomous digital entity a unique, cryptographically verifiable identity. Such a system would allow different platforms and services to recognize and trust agents from other providers, enabling interoperability. Without it, Cerf argues, we risk a fragmented ecosystem where every company builds its own walled garden of AI agents.
The problem is not theoretical. As AI agents become more common — from customer service bots to autonomous trading algorithms — the lack of identity standards could lead to fraud, confusion, and security breaches. A malicious agent could impersonate a legitimate one, or an agent could act beyond its authorization without a way to trace responsibility.
Creating a universal identity system for AI agents is technically complex. It requires agreement on cryptographic methods, governance structures, and liability frameworks. Cerf has argued that the internet's own development shows such challenges can
