A security researcher using Anthropic's Claude Opus 4.8 uncovered a critical flaw in Zcash's Orchard privacy pool in days, exposing a vulnerability that had survived four years of review by leading zero-knowledge cryptographers.
The bug could have allowed unlimited counterfeit ZEC creation. ZEC crashed roughly 38% on Thursday.
Ben Goertzel, CEO of SingularityNET, says frontier models are now reasoning about whether software behaves as intended, not just flagging coding mistakes. He calls this an early marker of a hard-to-overstate shift: security research is moving from slow, artisanal audits to continuous AI-driven review.
Sean Ren, CEO of Sahara AI and USC professor, warns that blockchain networks are especially exposed because their open-source code can be analyzed directly by frontier models, which can test attack strategies faster than traditional reviews. He notes that labs like OpenAI, Anthropic, and Google DeepMind have early access to unpublished models, creating a window for potential malicious use.
Danny Jenkins, CEO of ThreatLocker, says AI dramatically accelerates vulnerability discovery, broadening the pool of potential attackers beyond traditional experts.
Despite the risks, Goertzel says crypto may be better positioned than other industries to adapt due to its open code and security-focused communities.