Hackers aren’t breaking through corporate firewalls anymore. They’re walking right through the front door using your own voice. With over 10,000 AI voice cloning variants active online, traditional verbal verification is officially dead. Your identity has been turned into a free dataset for tech monopolies. This is the wake up call to start treating your voice like an asset. And our model is mapping this synthetic frontier natively as a security layer in businesses and finances sector.

Micro Update: The model had a clean week. We put it up against OpenAI, ElevenLabs, Azure, Voxtral and Deepgram, and not one miss across all five providers. Now adding more providers and we want you to decide what's next. Let us know which of the existing providers you most worried about.

Quick thing worth knowing as a first time or existing participants. We're not looking for clean runs. we're looking for where it struggles, where it gets confused or where it gets it wrong. That's the whole point of training this model. Break it if you can.

AI voice cloning is breaking one of the oldest forms of trust: the human voice. A year ago, most AI voice scams still sounded like isolated stories on the internet. Today, a few seconds of audio from a podcast, twitter space, tiktok, interview, or even a random voice note can be enough to clone someone convincingly. not just the voice itself. the pauses, emotion, tension in how they speak, the way they sound when they’re stressed or trying to explain something quickly. That’s how far this technology has already gone. The concerning part is that most people still think they’d be able to tell if a voice was fake. most can't, because this is becoming less of a “tech problem” and more of a human one. this is why conversations around AI detection matter now more than ever. We’ve been building toward this moment for a while. The Hyperdetexbot training is now open to people willing to contribute real voice samples and help improve detection systems moving forward.’ For now on the testnet there will be no data collection for training.