AI Uncovers Decade-Old Crypto Flaws

Crypto's security problems took a strange turn in 2026 when researchers using AI-assisted tools discovered several old flaws in various cryptocurrencies and wallets. One of the most significant findings was a four-year-old flaw in Zcash that could create counterfeit tokens without detection, which was only uncovered after researcher Taylor Hornby used Claude Opus 4.8 to audit the code.

The same AI tools were also used to find a weakness in Coldcard's firmware dating back to 2021, which had led to an estimated $100 million in bitcoin thefts across thousands of addresses. The researchers believe that unrestricted models were likely involved in these attacks, but attribution is still unclear.

Ai has also been implicated in other crypto-related incidents this year, including the manipulation of a Bankr AI agent into transferring around 3 billion DRB worth $150,000 to $200,000. Malicious on-chain dead-drop activity jumped by 440%, suggesting that powerful open-weight AI models are making it cheaper for attackers to build and deploy malware.

The researchers warn that crypto has years of public code waiting to be reread, and nobody knows how many old mistakes may still be hiding inside it. This raises concerns about the potential for future attacks and the need for better security measures in the crypto space.