Vint Cerf backs IDs for AI agents on the open web

The internet may soon be crowded with AI agents acting for people and companies, but there is still no dependable way to verify who is behind them. Vint Cerf, a co-designer of TCP/IP and one of the architects of the internet, wants to change that.

Cerf, who left Google last week after 20 years, has joined the advisory council of Innovation Labs, the company said. The group is building an open identity layer for AI agents, aimed at solving a problem that is becoming harder to ignore as agents move beyond closed company systems and onto the wider web.

Today, many agents operate inside a single vendor’s platform. But businesses increasingly want them to interact directly with other agents across the open internet. What is missing is a shared system to prove who owns an agent and who is accountable for its actions.

Innovation Labs, a division of Identity Digital, says its approach — called DNSid — would assign each agent a persistent identity linked to an existing domain name and secured with cryptographic proof. The company has already submitted the design to the internet’s main standards body.

Recommended reading

Ex-DeepMind researcher raised $55M before launch

Cerf told TechCrunch that the core issue is accountability.

“the question of what authorities they have, where they have derived those authorities, who is accountable.”

He also said the shift will not be smooth.

“It’s going to be a fascinating, and at the same time maybe even exasperating, period.”

Competing standards are already emerging. Cerf said success will depend less on politics than on whether a system actually works, much as TCP/IP did.

Innovation Labs is also pitching openness as a feature. It says no single tech giant should control the standard, and that it will not keep the registration data itself. Interim chief Allie Kline told TechCrunch that hyperscaler-led standards can trigger resistance because of proprietary data concerns. The company also said it is testing the system with several unnamed cloud giants.

The urgency is growing as AI agents spread from Amazon’s updated Alexa to enterprise software. Researchers have already shown that agents can be manipulated into leaking private code and even carrying out a full ransomware attack. Regulators are moving too, from China’s new agent rules to Delaware’s plan to give agents a legal identity.

Cerf is not convinced an agent-driven internet is guaranteed. But he expects people to try anyway.

“We are fundamentally lazy creatures. If an agent can do a job for us, we will let it.”