AI Hacking Hits Cryptocurrency..."Stole $100 Million by Exploiting Code Human Developers Missed for 5 Years"

▲ Cryptocurrency hacking, cryptocurrency crime/AI-generated image

Artificial intelligence (AI) is re-examining old blockchain code and continuously discovering critical security vulnerabilities that human developers have missed for years. As technological advancements create an environment for precise and low-cost analysis of vast open-source code, the speed race between security audits and hacking attacks is intensifying.

According to Bitcoin.com on September 20 (local time), Chainalysis reported a 440% surge in malicious on-chain command transmission activity, from an average of 2.06 incidents per day to 11.1 incidents. Similar to how the German Enigma cipher from World War II was deciphered after 85 years through cross-validation by AI agents in 1941, flaws in old source code that had been neglected for years in the virtual asset industry are now being continuously exposed through AI analysis.

A prime example is the discovery of a security vulnerability in the privacy coin Zcash (ZEC). Taylor Hornby, a researcher at Shielded Labs, used his own audit agent powered by Anthropic's AI model, Claude Opus 4.8, to identify a flaw in the Orchard circuit written in 2022. This vulnerability, which could have allowed the unlimited forging of Zcash, had been exposed for approximately four years but was swiftly patched thanks to AI analysis before any actual theft occurred.

Conversely, incidents have also occurred where security flaws were first exposed to hacker groups, leading to large-scale fund outflows. Coldcard, a hardware wallet manufactured by Coinkite, suffered the theft of over 1,600 BTC to 1,800 BTC, totaling more than $100 million in Bitcoin (BTC), from July 30 due to a random number generation flaw in its firmware distributed in 2021. The manufacturer concluded that the attackers likely used AI models during their review of the publicly available firmware code.

New attack techniques targeting vulnerabilities within AI models themselves have also emerged. The decentralized finance protocol Bankr was robbed of 3 billion DRB, worth approximately $150,000 to $200,000, when xAI's Grok misinterpreted malicious input written in Morse code as legitimate commands and executed them. This is an example of exploiting trust protocol flaws between machines to misuse AI as a hacking conduit.

With years of public code still accumulated across various blockchain networks, the competition to preemptively identify undiscovered flaws is expected to intensify. The market is keenly watching whether proactive adoption of AI-based security audit systems can defend against potential malicious threats in advance.

[Key Article Summary]

-Artificial intelligence (AI) re-analyzed blockchain code neglected for years, successively discovering Zcash (ZEC) forging flaws and Coldcard vulnerabilities.

-Coldcard suffered a theft of over 1,800 BTC, totaling more than $100 million in Bitcoin (BTC), due to a firmware flaw in 2021.

-As malicious on-chain activity surged by 440% according to Chainalysis, an AI security war to preemptively exploit code has begun in earnest.

*Disclaimer: This article is for investment reference only, and we are not responsible for any investment losses based on it. The content should be interpreted for informational purposes only.*