OpenZeppelin Co-Founder Manuel Aráoz Says He Considers “All” of DeFi Unsafe as AI Reshapes the Threat Model

Manuel Aráoz, former CTO and co-founder of blockchain security firm OpenZeppelin, said in an X post Tuesday that he now considers “all” of decentralized finance unsafe, citing the rise of AI coding agents as a structural threat that traditional audits cannot keep pace with.

“PSA: I now consider all of DeFi unsafe,” Aráoz wrote. “Coding agents are superhuman at finding vulnerabilities, and smart contract security is too asymmetric: defenders need to fix every bug while attackers need just one exploit to steal funds.” He added that he has advised friends and family to exit positions in major DeFi protocols including Aave, MakerDAO, and Compound, three of the most established lending and stablecoin platforms in the ecosystem.

This story is an excerpt from the Unchained Daily newsletter.

Aráoz’s standing matters even though he no longer represents OpenZeppelin. He co-founded the firm in 2015 alongside current CEO Demian Brener, and the OpenZeppelin smart contract library underpins audits for Aave, Compound, MakerDAO, Uniswap, Coinbase, and the Ethereum Foundation. He served as CTO before departing in 2019.

“Aráoz’s views do not represent OpenZeppelin‘s current position,” OpenZeppelin pushed back on the post publicly via a OpenZeppelin has been building AI-augmented security tooling, including a system called Skills that gives AI coding agents authoritative knowledge of audited smart contract libraries.

The numbers behind Aráoz’s warning are bleak.

DefiLlama data shows more than $1.1 billion lost to DeFi hacks over the past 365 days. April 2026 alone saw nearly $630 million drained across at least 27 reported exploits, the worst month for DeFi security since the Bybit incident in early 2025. The $292 million Kelp DAO bridge exploit on April 18, attributed to North Korea’s Lazarus Group, led the month, followed by a $285 million loss at Drift Protocol tied to a six-month social engineering campaign. Step Finance shut down earlier this year after a $27 million exploit it could not recover from. Since January 2026, more than $137 million has been drained from at least 15 DeFi platforms.

The structural argument Aráoz makes has not been resolved by the industry. Audits cost money, take weeks, and cover code as it exists at the time of review, not the version that a determined attacker probes weeks later. Anthropic has restricted public access to its Claude Mythos model in part because of concerns about its ability to autonomously discover and weaponize software flaws.

OpenZeppelin itself published a framework in May called the “Four Layers of DeFi Risk,” explicitly arguing that audits alone are no longer sufficient. The disagreement between Aráoz and his former firm is less about whether the threat has changed than about whether the right response is to retreat from DeFi entirely or to invest harder in AI-augmented defense.