Imagine if two AIs could chat with each other in a language that no human could understand. Right. Now go hide under the tires.
If you have called Customer Service in the last year or so, you’ve probably chatted with an AI. In fact, the earliest demonstrations of powerful large language models showed how such AIs could easily trick human callers. There are now so many AI chatbots out there that handle customer service that two of them are bound to call each other, and now, if they do, they can do it in their own special, sonic language.
Developers on Elevenlab’s 2025 Hackathon recently demonstrated Gibberlink. This is how it works according to a demonstration they delivered on YouTube.
Two AI agents from ElevenLabs (we have called them the best speech synthesis start -up) call each other for a hotel booking. When they are aware that they are both AI assistants, they change to a higher speed audio communication called GGWAVE. According to a post on Reddit, GGWAVE is “a communication protocol that enables data transfer via sound waves.”
In the video, the sound colors that replace spoken words sound a bit like old-school modem handshake protocols.
It’s hard to say whether GGWAVE and GIBBERLINK are faster than speech, but the developers claim GGWAVE is cheaper because it is no longer dependent on GPU to interpret the speech and can instead rely on the less resource -intensive CPU.
The group shared their code on GitHub if anyone wants to try to build this communication protocol for their own chat -Ai -chatbots.
Since these were Elevelab’s AI agents, there is no indication that Gibberlink would work with chatgpt or Google Gemini, though I am sure some will soon try similar GGWAVE efforts with these and other generative AI chatbots.
Look at
What do they say?!
A few artificial intelligence assistants “speak” their incomprehensible language sounds like a recipe for disaster. Who knows what these chatbots can get to? When they finish booking this hotel room, what if they decide to empty the user’s bank account and then use the funds to buy another computer to add a third GGWAVE “voice” to the mixture?
In the end, this is a Cool Tech demonstration that does not have much purpose besides proving that it can be done. However, people have managed to make people a little nervous.
