AI agents – Vinton Cerf, the 83-year-old co-inventor of the internet’s networking protocols, says AI is approaching a pivotal moment where openness, interoperable communication, and platform-building will decide whether agent-powered systems reach their potential.
Vinton Cerf has watched technology grow from a sketch into infrastructure—first the internet, now the AI boom. Speaking at the Open Frontiers conference this week alongside computer scientists and Databricks cofounder Matei Zaharia. the 83-year-old systems engineer didn’t treat AI’s next leap like a new mystery. He treated it like an old pattern showing up with different tools.
Cerf’s central message was blunt: AI is nearing an inflection point similar to the one that helped the internet become ubiquitous. At that moment, he said, the number of AI agents would make interoperability and standardization unavoidable.
He traced the internet’s early success to a simple decision about rules. “In the case of internet, it only worked because it was going to be distributed to begin with,” Cerf said. “And so we left the rules very open. We just said if you can find somebody to connect to and you follow the rules of the protocols. it should work.”.
That openness meant different kinds of networks could join the same conversation: a university network in California. a government research lab. and a commercial internet service provider could all connect by using the same technical language. Cerf framed AI’s current trajectory as heading toward a similar requirement—only now the “agents” that need to coordinate may multiply faster than any one system can handle.
The second challenge, Cerf said, will be how those agents communicate. Even if humans can speak to an agent in natural language, Cerf argued that everyday language may not be enough for agents to work together reliably or for developers to innovate without disrupting everything.
“I don’t think English is going to be the best choice,” Cerf said. “There is ambiguity, and I think precision for inter-agent interaction is going to be very, very important.”
He pointed to the way human language often depends on context, inference, and multiple meanings for the same words. For agents, he said, communication methods will have to minimize vagueness so each system can interpret requests and commitments consistently.
“An agent really needs to be sure the other agent understands what it is that they just agreed to do together,” Cerf added.
Cerf’s third lesson was about scale and staying power. The biggest technologies, he said, rarely remain standalone products. They become platforms—foundations that others can build on.
“A lot of the successes come from enabling technologies, whether it’s a platform or some other fundamental element that others can build on,” he said.
He offered familiar examples: Google, Amazon, Netflix, and millions of smaller developers built services on top of the same underlying internet infrastructure. Then he drew the parallel straight to AI.
“So if you really wanted to look for impact, think about things that enable other people to do things that they want to do,” Cerf said.
Taken together. Cerf’s guidance lands on the same set of pressure points: openness in the rules. precision in how agents coordinate. and the ability for transformative AI systems to serve as foundations rather than walled-off tools. For Cerf. the through-line is clear—when technology is ready to spread. the choices that shape interoperability and platform access decide who benefits when it finally scales.
Vinton Cerf Open Frontiers AI agents interoperability standardization internet protocols Databricks Matei Zaharia AI platforms open standards inter-agent communication
So basically we need AI to talk to each other like the internet did? k
Every time they say “standards” it just sounds like companies trying to lock people in. Like who even decides the rules, Google or the government? And isn’t AI already interoperable, or is that just marketing.
Wait I thought Vinton Cerf was the guy who made the internet protocols, right? So he’s saying AI agents will need to connect like websites? Idk, but if agents have to “follow rules” then doesn’t that mean they’ll all act the same and be boring? Also everyday language not enough… that’s kinda what Siri already does.
Openness is the whole thing though, like he says. But half the internet is already broken with all the “walled gardens” so I’m skeptical. People keep building chatbots that can’t even share info, so how is this any different? Also I read somewhere AI agents will “multiply” and then we’ll all be stuck paying subscriptions to connect them… seems like that’s what they mean by standards.