Frontier AI models such as Anthropic's Claude Opus 4.8 and OpenAI's GPT-5.5 are increasingly being used to identify software vulnerabilities, raising security concerns. This week, Zcash developers disclosed that Claude Opus 4.8 helped discover a critical vulnerability in the Orchard privacy pool that could have allowed attackers to mint unlimited ZEC. Due to the privacy design, it's impossible to determine if exploitation occurred, causing ZEC's price to crash. Experts warn that as AI tools become more capable and accessible, the pace of vulnerability discovery will accelerate, posing risks to open-source crypto projects. While some advocate restricting AI access, others argue for democratizing defensive tools to avoid an imbalance between attackers and defenders. The trend is already impacting DeFi, with over $840 million stolen in early 2026, partly attributed to AI-assisted attacks.

Key facts

• Claude Opus 4.8 helped discover a critical Zcash bug enabling unlimited ZEC minting.
• ZEC price crashed after disclosure; exploitation status remains unknown due to privacy.
• AI models like GPT-5.5 are accelerating vulnerability discovery across software.
• Debate arises on restricting AI access vs. democratizing defensive tools.
• Over $840 million stolen from DeFi in early 2026, partly from AI-assisted attacks.