AI-driven attacks cause record DeFi losses in April 2026
DeFi protocols suffered approximately $722.4 million in losses in April 2026 from cyberattacks, with reports linking some attacks to North Korean actors and others to unidentified threat actors. Security experts warn that AI systems like Anthropic's Claude Mythos can detect software vulnerabilities faster than human defenders, creating a significant asymmetry in offense versus defense in the DeFi ecosystem.
What happened
In April 2026, decentralized finance protocols suffered approximately $722.4 million in losses from attacks, according to Dune Analytics. Protocols lost around $285 million in attacks reportedly linked to North Korean actors, with losses from unidentified parties totaling $437.4 million.
Most incidents in 2026 stemmed from vulnerabilities in bridges and social engineering attacks. Manuel Araoz, founder of OpenZeppelin, warned that AI-powered software can scan and find vulnerabilities much faster than conventional methods, creating an uneven playing field between offense and defense.
Reported impact
According to DeFiLlama, the total value locked in DeFi platforms has dropped by $20 billion since the beginning of the year. Over $1.1 billion in assets have been lost to various security incidents within the DeFi ecosystem in the past year.
Notable attacks included the April 2024 Kelp DAO hack resulting in a $292 million loss, and the Solana-based Step Finance platform being forced to shut down entirely after a $27 million hack earlier in 2026.
Index analysis
Araoz emphasized that AI-based agents are far more skilled than humans at detecting weaknesses in software, and while defenders must fix every bug, attackers only need to find one to steal funds. Anthropic's Claude Mythos is described as an AI model capable of independently locating software vulnerabilities and instantly generating working exploit code.
Because smart contract code is publicly visible, AI is able to detect flaws faster than humans and instantly launch attacks. The foundation of DeFi's current security models was built to counter threats from human actors, not sophisticated AI systems.
Governance controls
Some investors are being encouraged to avoid 'permissionless transactions' and adopt time lock mechanisms as added precautions. Araoz declared that he currently finds 'all' DeFi applications untrustworthy due to new risks posed by AI agents capable of analyzing code at a 'superhuman' level.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.