‘All Of DeFi Unsafe,’ Developer Warns As AI Agents Reshape Security Threats

Attackers drained an estimated $200,000 from DeFi liquidity pools on Ethereum — specifically Uniswap V3 — after exploiting weaknesses in the WUSD.fi and GLOVE incentive system, according to security researchers at ExVul.

The attackers cycled funds through multiple wallets to repeatedly farm rewards, taking advantage of flaws baked into the protocol’s incentive structure.

A Wave Of Attacks Hitting The Ecosystem

That incident was one of several to rock the DeFi space in recent days. Fraudulent Google advertisements impersonating Uniswap also surfaced earlier this week, routing unsuspecting users to phishing sites designed to steal wallet credentials — a scam that reports say drained at least $400,000 before it was flagged.

The back-to-back incidents set the stage for a blunt public warning from Manuel Aráoz, the founder of OpenZeppelin, one of the most widely used smart contract security firms in the industry.

Aráoz said he now considers all of DeFi unsafe, a statement that spread quickly across developer circles after he posted it online.

His reasoning cuts to a basic problem in how blockchain security works. Defenders have to find and patch every single vulnerability, while an attacker only needs one to drain a protocol entirely.

AI Tools Shifting The Balance

Aráoz pointed to AI-powered coding tools as the reason that balance has gotten harder to manage. Reports indicate he believes these tools allow attackers to scan contracts for weaknesses at a speed and scale that most security teams cannot match.

He went further in private communications, reportedly advising friends and family to pull their funds from major DeFi platforms altogether, including Aave, MakerDAO, and Compound. Those three platforms represent a significant share of total value locked across decentralized finance.

Cybersecurity analysts have raised similar concerns, warning that AI is accelerating how fast attackers can map out vulnerabilities, build phishing infrastructure, and run simulated exploit strategies against live protocols.

The problem is compounded by how modern DeFi protocols are built. Many now stack multiple components on top of each other — bridges, lending systems, staking mechanisms, automated reward contracts — and each additional layer widens the surface area that has to be defended.

OpenZeppelin itself previously flagged how dangerous these combinations can be, identifying a vulnerability that emerged from the interaction between ERC-2771 and Multicall standards, two widely used contract types that created unintended exposure when used together.

Major protocols have responded by pouring resources into audits, bug bounty programs, and formal verification. Reports note that even those efforts have not fully closed the door on phishing attacks and incentive manipulation schemes.

The concern now is whether smaller DeFi projects — those without the budget for continuous security reviews — can hold up against attackers who are moving faster than before.

Featured image from Binance, chart from TradingView