Moonwell Hit by $1.78M Exploit as AI Vibration Coding Debate Reaches DeFi

Featured in:
abcd

Moonwell, a decentralized finance (DeFi) lending protocol deployed on Base and Optimism, was exploited for approximately $1.78 million after the price oracle for Coinbase Wrapped Staked ETH (cbETH) returned a value of approximately $1.12 instead of $2,200, creating a mispricing that attackers could exploit for profit.

Moonwell he said following a post-mortem that found that the governance proposal made on Sunday misconfigured the cbETH oracle using only the cbETH/ETH exchange rate, which caused the system to report cbETH at around $1.12. The minutes said liquidation bots and opportunistic borrowers took advantage of the mispricing, leaving approximately $1.78 million in bad debt.

sadasda

The pull requests affected contracts show multiple commits co-authored by Anthropic’s Claude Opus 4.6, prompting security auditor Pashov to go public flag incident as an example of the backfire of AI-written or AI-powered Solidity.

Speaking to Cointelegraph about the incident, he said he linked the case to Claude because there were multiple commits in the pull requests Claude co-authored, meaning that “the developer was using Claude to write code, which led to the vulnerability.”

However, Pashov cautioned against treating this vulnerability as solely based on artificial intelligence. He described the Oracle issue as a mistake that “even a senior Solidity developer could have made,” arguing that the real problem was a lack of sufficiently exacting checks and comprehensive validation.

A code vulnerability led to the Moonwell exploit. Source: Paszov

He initially stated that he believed no testing or audit had been performed at all, but later admitted that the team had stated that they had placed the unit and integration tests in a separate pull request and had outsourced the audit to Halborn.

He said the mispricing “could have been detected by an integration test, a proper one, involving integration with the blockchain,” but he declined to directly criticize other security companies.

Related: How South Korea uses artificial intelligence to detect manipulation in the cryptocurrency market

Small loss, substantial management questions

The dollar value of the exploit is minuscule compared to some of the biggest DeFi incidents, such as the March 2022 Ronin Bridge exploit in which attackers stole over $600 million, or other nine-figure breaches of bridges and lending protocols.

What sets Moonwell apart is a combination of AI co-authorship, a seemingly uncomplicated error in the price configuration of a major resource, and existing audits and tests that failed to detect it.

Pashov said his own company wouldn’t fundamentally change its process, but if the code appeared to be “vibrate-coded,” his team would “keep their eyes a little wider open” and expect a higher density of low-hanging problems, even though this particular oracle bug “wasn’t that easy” to detect.

“Vibe coding” and the disciplined utilize of artificial intelligence

Fraser Edwards, co-founder and CEO of decentralized identity infrastructure provider cheqd, told Cointelegraph that the vibration coding debate masks “two very different interpretations” of how AI will be used.

Related: How AI cryptocurrency trading will make and break human roles

On the one hand, he said, non-technical founders encourage AI to generate code they can’t independently check; on the other hand, experienced developers using AI to accelerate refactoring, pattern mining and testing as part of a mature engineering process.

AI-powered development “can be valuable, especially for MVPs [minimal viable product] stage,” he noted, but “should not be considered a shortcut to production-ready infrastructure,” especially in capital-intensive systems like DeFi.

Edwards argued that all AI-generated sharp contract code should be treated as untrusted input, subject to strict version control, clear code ownership, multi-person peer review and advanced testing, particularly in high-risk areas such as access control, Oracle and pricing logic, and update mechanisms.

“Ultimately, responsible AI integration comes down to governance and discipline,” he said, specifying clear review gates, separation of code generation from validation, and the assumption that any contract implemented in an adversarial environment may have hidden risks.

Warehouse: South Korea is getting affluent on cryptocurrencies… North Korea is acquiring weapons

Cointelegraph is committed to independent and see-through journalism. This news article has been produced in accordance with Cointelegraph’s Editorial Policy and is intended to provide exact and up-to-date information. Readers are encouraged to verify the information themselves. Read our Editorial Policy https://cointelegraph.com/editorial-policy
abcd
sadasda

Find us on

Latest articles

Related articles

See more articles

SocGen’s FORGE extends euro stablecoin to XRP Ledger in...

The digital assets arm of French banking group Societe Generale, SG-FORGE, has launched its euro-denominated stablecoin, EUR...

Bitcoin could benefit if AI job losses cause bank...

Arthur Hayes created a raw market warning: Sees the growing divide between his preferred risk measure, Bitcoin,...

XRP is emerging as the cryptocurrency everyone is talking...

They say journalists are never really done. But for Christian it's not just a metaphor, it's a...

Peter Thiel’s founding fund abandons stake in ETHZilla amid...

Billionaire Peter Thiel's founding fund has completely exited treasury company ETHZilla, according to a Tuesday filing with...

Is Bitcoin supply heading to mighty hands? Whale data...

Bitcoin continues to struggle to regain the $70,000 level, with continued selling pressure keeping the market on...

Bitcoin tech stock divergence is a ‘fire alarm’ for...

The divergence between Bitcoin and tech stocks is a warning sign of a potential AI-driven credit crisis...