AI Coding Agents Have Made All DeFi Unsafe, Security Expert Says

Critics rapidly pushed again, arguing that almost all current crypto hacks had been attributable to operational failures as an alternative.

Manuel Aráoz, co-founder of sensible contract safety agency OpenZeppelin, went public on Could 26 with a blunt advice that folks ought to get out of DeFi, all of it, together with the blue chips.

In accordance with him, AI-powered coding brokers have tilted the safety recreation to this point towards attackers that no protocol can presently be trusted to carry consumer funds.

Aráoz’s Warning

The software program engineer wrote in a put up on X;

“PSA: I now contemplate all of DeFi unsafe.”

He additionally mentioned he has been privately advising family and friends to exit all DeFi positions, naming Aave, MakerDAO, and Compound as protocols he now not considers secure.

His reasoning is predicated on asymmetry: defenders should discover and repair each vulnerability, whereas attackers want just one to trigger harm. Now, with AI coding brokers able to scanning sensible contracts quicker and extra completely than any human safety staff can, Aráoz feels the asymmetry has grow to be unworkable.

OpenZeppelin itself lately noted that crypto corporations misplaced greater than $3.4 billion to hacks in 2025; nonetheless, it blamed most of that theft on compromised credentials, operational failures, and code shipped between audits, slightly than on sensible contract bugs.

This 12 months has additionally seen a rollercoaster of assaults, with greater than $650 million stolen in April alone. Of that quantity, $292 million got here from an exploit on KelpDAO, with one other $285 million siphoned from Drift Protocol following what specialists say had been months of social engineering.

You may additionally like:

Pushback From X Customers

Towards that backdrop, Aráoz’s warning landed arduous, however folks instantly pushed again. A kind of criticizing the put up was Aave Chan Initiative founder Mark Zeller, who held nothing again.

His counter was data-driven: he pointed out that fewer than 10% of DeFi points prior to now 12 months stemmed from code-level vulnerabilities, with most failures, in line with him, tracing again to poor danger parameters, collateral mismanagement, and weak operational safety, not AI-assisted exploits.

A number of others echoed Zeller’s view, although with barely much less warmth. Phoenix Lab co-founder Sam McPherson indicated that sensible contracts of blue-chip DeFi platforms had been “fairly secure lately” and pointed to opsec failures as the true offender behind a lot of the main hacks which have occurred lately.

One other X consumer, Polaris Finance developer Robert, made the same distinction, saying that precise sensible contract exploits are “virtually non-existent lately.” He added that current breaches have largely concerned centralized elements that enable human management slightly than the immutable code beneath them.

Ethereum co-founder Vitalik Buterin additionally has a special view on AI and its impact on crypto safety, writing earlier this month that AI-assisted formal verification may truly make crypto programs safer over time. In accordance with him, builders can use AI to write down each the code and the mathematical proofs of its correctness.