Vint Cerf, one of the principal architects of the internet's foundational protocols, has turned his attention to a new challenge: giving artificial intelligence agents verifiable identities as they begin to operate autonomously across the open web. Having left Google after two decades, Cerf is now advising Innovation Labs, a subsidiary of domain registry company Identity Digital, on a proposed standard called DNSid.
The system would assign each AI agent a cryptographic identity linked to an existing internet domain name, creating a permanent, auditable record of its registration and the organisation responsible for it. According to Allie Kline, interim CEO of Innovation Labs, the standard is already being tested with several major cloud providers and identity firms, though their names have not been disclosed.
Cerf told TechCrunch that the initiative addresses fundamental questions about authority, accountability and trust in an era where software agents may increasingly act on behalf of people and businesses. 'What authorities do they have, where have they derived those authorities, who is accountable for the behaviour of an agent?' he said. He noted that agents are far more dynamic than domain names, making the commitment behind each registration unclear.
The push for a shared identification standard comes as UK businesses explore using AI agents for tasks ranging from automated customer service to supply chain coordination. Currently, most agents operate inside proprietary systems and cannot interact with agents built by other companies. The UK's Information Commissioner's Office has emphasised that organisations must ensure AI systems are transparent and accountable under data protection law. Meanwhile, the EU AI Act, which came into force earlier this year, imposes strict requirements on high-risk AI systems, including traceability and human oversight.
Cerf acknowledged that multiple competing standards are emerging, but argued that user demand for interoperability will ultimately drive adoption, just as it did for TCP/IP. 'Company X uses agent Y's technology, and company A uses agent C's technology, and then they don't interwork with each other,' he said. 'Nobody can do everything that you might want every agent to do.' Kline stressed that Innovation Labs does not plan to own the registration data, hoping to avoid the 'organ rejection' that would greet a proprietary standard from a big tech firm.
For UK consumers, the prospect of AI agents interacting on their behalf raises both convenience and security concerns. Without a reliable way to verify an agent's identity, scams and data misuse could proliferate. A universal identification layer could help consumers and businesses trust that an agent is who it claims to be, and hold its operator accountable if something goes wrong. Cerf was cautious about whether such an 'agentic economy' is inevitable, but said: 'We are fundamentally lazy creatures, and if we find a way to have an agent do something for us, we're very likely to choose to do that because it's just easier.'