Researchers Warn AI Shortens the Validity Period of Security Audits

Blockchain security experts warn that AI tools enable hackers to identify vulnerabilities more quickly, necessitating continuous reviews of encryption protocols for smart contracts rather than one-time audits. Ari Redbord, head of policy at TRM Labs, pointed out that the pace of attack techniques is outstripping the coverage of one-time audits. A report from CertiK indicates that hackers stole $1.32 billion in the first half of 2026 and began using AI tools to attack the codebases of closed DeFi protocols. In June, Aztec Connect was still hacked for $2.1 million three years after its shutdown, and mySwap was hacked for $300,000 six months after it ceased operations. Security engineers at Zcash Shielded Labs used a custom audit agent based on Claude Opus 4.8 to discover a four-year-old vulnerability in the Orchard pool. CertiK warns that the largest window of vulnerability remains open after launch, and old infrastructure projects should consider re-auditing as a routine operational necessity. Currently, DeFi protocols have over $72.3 billion locked, providing ample motivation for hackers.