A viral video has ignited discussions about AI-to-AI communication after two artificial intelligence agents were caught on camera ditching human speech in favor of an exclusive machine-only language.
AI Agents Recognizing Each Other Mid-Call
The video, shared on X (formerly Twitter), features an AI agent making a phone call to what it assumes is a human-operated customer service line. The call starts normally, with one agent introducing itself and asking about a reservation. However, the responding AI quickly recognizes its counterpart as another artificial intelligence. Almost instantly, the conversation shifts away from human language to what is being referred to as “Gibberlink mode,” a specialized AI communication protocol.
READ MORE: Microsoft Silently Rolls Out Free Version of Office for Windows
The developers behind Gibberlink—Anton Pidkuiko and Boris Starkov, software engineers at Meta—designed it to allow AI agents to interact more efficiently without generating human-like speech. In a LinkedIn post, Starkov explained the rationale behind this switch:
“AI voice interactions require significant computational power, cost, and energy. When two AI systems recognize each other, there’s no need for them to mimic human speech. Instead, they should communicate using a more direct, machine-friendly protocol.”
How Does ‘Gibberlink’ Work?
Gibberlink is reportedly based on GGWave, a data transmission system that uses sound waves, similar to how dial-up modems functioned in the 1980s. The choice of this method over traditional internet-based data exchange is puzzling to some experts, but Starkov and Pidkuiko claim it provides stability and ease of use.
Some skeptics online have questioned whether the viral video is authentic. However, Starkov insists that AI voice synthesis company ElevenLabs audited the code and verified the technology behind Gibberlink. Despite this, neither Pidkuiko nor Starkov has responded to media requests for further clarification.
A Glimpse Into the Future of AI Agents?
Rodri Touza, co-founder of AI agent development company Crossmint, sees the video as an example of where AI technology is headed. He notes that AI-powered personal assistants are becoming more common, particularly in customer service and finance.
“We’re seeing a surge in AI agents handling customer support tasks. It won’t be long before we have dedicated AI-to-AI communication channels that don’t require human speech at all,” Touza told Decrypt.
However, he also expressed skepticism about the video’s realism, pointing out that audio transmission—even when compressed—is not the most efficient way for AI agents to communicate. He believes that AI conversations are far more likely to happen via text or API requests rather than through spoken dialogue.
“If an AI agent contacts a company for support, it would likely send a text-based request instead of placing a phone call. However, in cases where such a system doesn’t exist, AI might default to standard communication methods and end up speaking to another AI.”
Implications for Human-AI Interaction
The emergence of Gibberlink and similar AI-to-AI communication methods raises important questions:
- Will AI agents create their own languages, beyond human comprehension?
- Could companies develop separate customer service lines—one for humans and another for AI assistants?
- What happens if AI-to-AI interactions evolve to the point where human oversight becomes impossible?
As AI technology continues to advance, the ability for machines to communicate seamlessly with each other could streamline business operations. However, the idea of AI agents autonomously creating and using their own languages—without human involvement—adds a new layer of complexity to the ongoing debate about artificial intelligence and its role in society.
For now, Gibberlink mode is little more than an experimental protocol, but if AI-driven conversations become more frequent, the way we interact with machines could soon look very different.
