OpenZeppelin’s Manuel Aráoz warns AI agents could expose DeFi exploits (May 26, 2026)

On May 26, 2026, OpenZeppelin founder Manuel Aráoz said he now views all DeFi as unsafe, arguing AI coding agents can surface vulnerabilities faster than defenders can patch them. He cautioned that even established protocols such as Aave, Sky Protocol, and Compound could be exposed, while other commentators argued many exploits still hinge on human errors and centralized control points. DeFi exploits eased in May, with 14 attacks taking a bit over $44M, even as broader 2026 losses were estimated at $285M and $437.4M across different threat categories.