DeFi security is facing new challenges as automated code review and agentic tooling become increasingly prevalent. According to DefiLlama data, over $1.1 billion was lost to DeFi exploits between June 2025 to June 2026.
OpenZeppelin co-founder Manuel Aráoz warned that coding agents are becoming 'superhuman' at finding vulnerabilities, which could give attackers a faster way to scan for exploitable weaknesses.
A recent Zcash case showed the other side of the AI security debate. A researcher used a frontier model to identify a critical four-year-old Orchard bug, highlighting how AI can surface issues that remain hidden across earlier review cycles.