$2 AI Model Quickly Identifies ColdCard Vulnerability

An AI model called GLM 5.2 reportedly rediscovered the vulnerability behind a recent ColdCard theft in approximately 20 minutes for about $2.

The test, which took place after the incident, does not establish that AI was involved in the theft.

The low-cost model can be used to review old wallet code and cryptographic implementations repeatedly, making it easier for attackers and developers to examine code protecting large amounts of cryptocurrency.

Affected users had followed standard security advice, storing their Bitcoin in dedicated hardware wallets. However, self-custody still relies on the device creating and protecting its keys correctly.