OpenZeppelin co‑founder warns all DeFi is unsafe amid AI‑driven

Manuel Aráoz, a co‑founder of OpenZeppelin, announced on X that he now considers the entire decentralized finance (DeFi) sector unsafe because AI‑powered coding agents can locate vulnerabilities faster than defenders can patch them [1]. He privately advised friends and family to withdraw from even the most established protocols such as Aave, MakerDAO and Compound.

Key takeaways

• Aráoz cites superhuman AI agents as the primary reason DeFi security is now “asymmetric” [1].
• Recent data shows over $1.1 billion lost to DeFi hacks in the past year, with April alone accounting for $292 million [2].
• Critics argue that less than 10 % of DeFi incidents stem from smart‑contract code, labeling Aráoz’s warning as fear‑marketing [1].
• OpenZeppelin clarified that Aráoz left the company in 2019 and his views do not represent the firm’s current stance [1].
• Industry figures suggest AI tools can also bolster defenses, prompting calls for circuit breakers and timelocks [1].

AI‑driven vulnerability hunting reshapes DeFi risk

Anthropic’s release of the Mythos model earlier this year marked a turning point, according to Aráoz. The model, restricted to a small partner group, has already uncovered critical bugs in long‑running software that escaped human detection [1]. Exchanges such as Coinbase have reportedly reached out to Anthropic for access, underscoring the perceived threat. A high‑profile $120 million exploit last year demonstrated how a contract that survived multiple audits could still be compromised, echoing the “penny‑skimming” scenario from the film Office Space [1].

The surge in attacks continued into 2026, with April recorded as the worst month for crypto hacks, averaging nearly one incident per day [1]. A notable recent breach involved stablecoin issuer StablR, where an attacker seized a single key in a 1‑of‑3 multisig wallet, minted $13.5 million of unbacked tokens, and swapped them for roughly 1,115 ether (about $3 million) [1]. These incidents highlight that many exploits arise from social engineering and poor operational security rather than pure code flaws [1].

Community response and divergent views

Aráoz’s statement sparked sharp disagreement. Marc Zeller of the Aave Chan Initiative dismissed the claim as “moronic,” noting that under 10 % of DeFi problems in the past year were due to the codebase itself [1]. Some observers suggested the warning serves OpenZeppelin’s own marketing interests, though Aráoz emphasized his concern spans parameters, mechanism design, and operational security [1]. OpenZeppelin publicly distanced itself, stating his comments do not reflect the company’s current position under CEO Demian Brener [1].

Conversely, industry voices such as Alchemy’s Uttam Singh advocate for “gated DeFi” measures—circuit breakers, timelocks, and security councils—to mitigate the asymmetry until the sector matures [1]. Founders Hayden Adams () and Stani Kulechov (Aave) countered that the same AI tools can be leveraged by defenders, arguing that DeFi’s security landscape is evolving rather than collapsing [1].