Why a DeFi Security Insider Just Told People to Get Out



One of the people most responsible for making DeFi's largest protocols auditable just said he no longer trusts them.

Manuel Aráoz, co-founder of OpenZeppelin — the security firm behind audits for Aave, Compound, MakerDAO, and Uniswap — posted publicly on May 26 that he now considers all of DeFi unsafe. He said he had already privately advised friends and family to exit positions before making the statement public.

His reasoning is not emotional. It is structural.

Smart contract security has always worked on an asymmetric basis. Defenders must close every gap. Attackers only need one. That was always true. What Aráoz says has changed is that AI coding agents are now operating at superhuman speed when scanning code for exploits. The time advantage defenders once had is narrowing or gone.

This is not a fringe view showing up from outside the industry. OpenZeppelin is the firm that helped build the security standards the entire sector relies on. When someone with that background says the model is broken, it carries weight that most external warnings do not.

The numbers in 2026 support the concern. More than $600 million was drained from DeFi protocols in April alone. KelpDAO, Drift, and Euler all suffered major losses within the same month. The scale and frequency of these incidents have been rising, not falling.

Not everyone agrees with Aráoz's framing. Paloma Chain pushed back on X, arguing that KelpDAO's loss was an infrastructure and operations failure, not a smart contract exploit. That distinction matters for how risk is categorized. But it does not change the broader direction of the debate.

For anyone holding positions in yield protocols, lending pools, or liquidity provision right now, this conversation is directly relevant. The question is not whether AI changes the threat environment. It is how fast and whether current audit and monitoring practices can keep up.

The full breakdown of the debate, the pushback from Morpho, the legal questions raised about Aráoz's pre-public advisory to friends, and what on-chain data from April 2026 actually shows is covered in detail at CryptoNewsLive.org.

Read the full analysis here: CryptoNewsLive.org 

Comments

Post a Comment

Popular posts from this blog

Ripple Is Building XRPL's Defense Against Quantum Computing, and the Clock Is Already Running

Image
  Ripple has released a detailed plan to protect the XRP Ledger from quantum computing threats, and the timeline it has set is not aspirational. The target is full post-quantum cryptography readiness by 2028, with active work already underway in 2026. The roadmap runs four phases. The first is an emergency contingency for a scenario where quantum computers break classical cryptography ahead of schedule. Ripple calls this Q-Day readiness. In that scenario, the XRP Ledger would enforce a hard network shift, stopping classical public-key signatures and requiring users to migrate to quantum-safe accounts. The second and third phases cover 2026. Testing begins on NIST-recommended post-quantum algorithms under real XRPL workload conditions. Performance, storage, and bandwidth costs all get measured. Then hybrid deployment begins on Devnet, running quantum-resistant signatures alongside existing ones before anything touches the live network. Project Eleven, an organization working on cr...
Read more

Hoskinson Just Said Everything Nobody Else Will Say About Crypto in 2026

Image
  Charles Hoskinson spent over an hour on the O Show with host Wendo in April 2026 and covered ground that most crypto founders avoid in public interviews. The Midnight Network mainnet went live on March 31 with nine institutional node operators including Google Cloud, MoneyGram, Worldpay, Telegram, Vodafone's Pairpoint, eToro, Bullish, Blockdaemon, and Shielded Technologies. That alone is a headline. But it was not the most important thing he said. The AI agent argument is where this interview breaks from the usual. His position is that zero-knowledge proofs are not primarily a privacy feature. They are the only mechanism by which AI agents can verify each other's behavior. When an autonomous agent is managing your funds across multiple blockchains while you sleep, it needs mathematical proof infrastructure underneath it to constrain what it can do, protect your data without exposing it, and settle value in real time without a middleman. Midnight is being built for that future...
Read more

Everclear Is Gone and the $500M Volume Story Should Worry Every DeFi User

Image
 Everclear, the cross-chain settlement protocol that processed $500M in monthly volume, announced this week that it is shutting down. The Foundation and Labs entities are winding down operations. The product UI is already offline. The chain is non-operational. This is not a rug pull. It is something arguably more instructive: a protocol that built real volume, real technology, and real partnerships, and still could not survive. The team's own explanation makes it plain. Users were highly price-sensitive. The solver-based model that Everclear was built around never developed the revenue depth it needed, even with hundreds of millions of dollars flowing through the system monthly. A B2B2C pivot was attempted. Major industry names signed on. But the timelines to go live were longer than the runway the team had left. Acquisition talks were explored. They did not succeed. The CLEAR token has lost more than 52% of its value in seven days. A potential buyback is being explored in the ...
Read more