Hackers are using advanced AI tools to break into deployed smart contracts, rendering traditional one-time security audits obsolete. In the first half of 2026 alone, attackers drained $1.32 billion from the Web3 ecosystem.
The rapid adoption of generative AI allows malicious actors to scan complex, closed-source code repositories in minutes. That process used to take human engineers weeks to complete manually. Now, the window between vulnerability discovery and execution is almost non-existent.
The End of the “One and Done” Era
CertiK confirmed the $1.32 billion loss in its H1 2026 Hack3D report. The numbers show a sharp reversal of recent industry progress. DeFi hacking losses had dropped from a peak of $2.62 billion in 2022 down to roughly $680 million in 2025. The sudden jump back over the billion-dollar mark this year points directly to the speed of AI-accelerated exploits.
Even abandoned projects are getting hit. Hackers exploited Aztec Connect for $2.1 million three years after the protocol officially closed its doors.
Another platform, mySwap, lost $300,000 six months after shutting down, according to a detailed report by Phemex that also documented Zcash’s recent use of Claude Opus 4.8 for vulnerability discovery. The ability of AI models to retroactively analyze old code has turned inactive protocols into easy targets.
Fighting Fire With Fire
Developers are fighting back using the exact same tools. Engineers at Zcash Shielded Labs recently deployed a custom auditing agent powered by Anthropic’s Claude Opus 4.8.
The AI successfully unearthed a hidden, four-year-old vulnerability buried within the Zcash Orchard pool. That catch saved the network from a potential disaster. And it proved that LLMs are just as capable of defending networks as they are at breaking them.
Why Continuous AI Monitoring Is the New Standard
Security firms are urging the $70 billion DeFi sector to abandon the old way of doing things. A single, pre-launch manual audit used to be the gold standard. That is no longer true.
Protocols must now treat security as a continuous operational expense. Regular re-audits and active AI monitoring are becoming strict requirements to survive. In this new arms race, the shelf life of a clean audit can expire overnight.