CoinDesk reported that recent major security incidents involving Bitcoin infrastructure have raised concerns that artificial intelligence (AI) could significantly lower the cost of discovering deep vulnerabilities in financial software. These incidents include the theft of approximately $114 million in Bitcoin from a Coldcard wallet, Core Lightning developers issuing an emergency patch after an AI-generated security report identified a vulnerability, and Blockstream's Liquid Network being exploited by a white-hat hacker, leading to the extraction of about 4,000 Bitcoin (worth approximately $317 million), although 3,400 of them have since been returned.

These vulnerabilities highlight a paradox in Bitcoin's design: the main layer is deliberately kept simple to minimize risk, but smart contracts and off-chain scaling layers (Layer-2s) developed to introduce more utility and speed bring more complex and potentially more vulnerable codebases. Furthermore, AI is being used extensively for vulnerability hunting. In August, 16 Bitcoin developers used AI models to scan 390 Bitcoin projects, finding nearly 5,000 issues, 85 of which were rated as critical. Gregory, a Bitcoin application developer, noted that AI is finding vulnerabilities that humans cannot, and it has changed the risk paradigm for older financial software; even unused old code could pose a renewed threat due to the reduced cost of AI reading it.