Prem AI launches CyberScan for continuous AI security audits of Bitcoin infrastructure

London, UK, August 20th, 2026, Chainwire

CyberScan turns frontier AI models into repeatable, resumable security reviews that deliver actionable findings inside developers’ existing workflows.

Prem AI, the Sovereign Intelligence company with R&D and operations in Switzerland, today launched CyberScan, a continuous AI security audit harness built initially for Bitcoin infrastructure. Leading Bitcoin companies ArkLabs and Breez have had early access to CyberScan during the beta, running it against production code ahead of launch. 

CyberScan gives engineering teams an ongoing layer of code review between formal security audits. It repeatedly analyses repositories as code evolves, preserves scan state if a run is interrupted, and returns prioritised findings with severity ratings and file-level references. It is designed to complement expert review, not replace it.

CyberScan’s launch follows the year's largest hardware wallet exploit. On July 30, an attacker pulled more than $116 million in bitcoin from over 5,200 Coldcard hardware wallets, in four waves over a single week, according to TRM Labs.

The root cause was a firmware regression introduced in March 2021; human reviews on the open-source repository missed it, and automated AI attacks leveraged it: a build-time check meant to guarantee the wallet used its hardware random number generator instead silently let it fall back to a deterministic software PRNG. Every wallet seed became predictable, and every human review missed it for five years. The attacker rebuilt private keys offline, without ever touching a device.