AI Found Zcash's Four-Year Flaw Before Criminals Did

The gap between defenders and attackers using AI is shrinking. Hornby's discovery required a custom framework and a single model. The cost of replicating that effort for malicious purposes is dropping fast.

Anthropic restricts access to its most advanced Mythos models to vetted security teams. The company said Mythos-class models can "discover and exploit software vulnerabilities and run portions of agentic cyber operations on their own." After the Zcash fix, Zcash founder Zooko Wilcox said a Mythos audit "did not find any more serious bugs."

A clean audit holds only until the next generation of models ships. ZEC holders saw the token drop 60% in the months after the disclosure, despite the flaw never being exploited.

The NU6.2 hard fork activated on June 3, closing the Orchard vulnerability.