​Claude Opus 4.6 linked to $1.78 mln crypto exploit

​Claude Opus 4.6 linked to $1.78 mln crypto exploit
Claude Opus 4.6 at center of $1.78 mln Moonwell exploit

Anthropic has launched next generation of its models — Claude Opus 4.6 and Sonnet 4.6 — featuring context windows of up to 200,000 tokens, adaptive thinking mode and enhanced coding tools. However, Claude Opus 4.6 has already been drawn into controversy after DeFi protocol Moonwell suffered $1.78 mln in losses due to a smart contract error.

Company positions Claude 4.6 as model for agent building and software development. Opus 4.6 supports up to 128,000 output tokens, adaptive thinking mode, free code execution when paired with web tools and new dynamic filtering capabilities.

On Feb. 15, Moonwell, which operates on Base and Optimism, incurred losses following oracle update under governance proposal MIP-X43. Part of code was generated or edited using Claude Opus 4.6, Cryptopolitan reported.

cbETH mispricing triggered liquidations

Due to error, Coinbase Wrapped Ether (cbETH) was valued at about $1.12 instead of its actual market price near $2,200. Code retrieved cbETH/ETH exchange rate but treated it as if it were denominated in US dollars.

This led to mass liquidations, allowing users to repay roughly $1 in debt and receive collateral worth thousands. Protocol quickly reduced borrow cap to 0.01, but some losses had already materialized. Total damage was estimated at $1.78 mln.

Moonwell said issue was isolated to cbETH Core Market on Base, with no impact on other markets.

Vibe coding debate intensifies

Incident has reignited debate over use of AI in software development. Auditors stressed responsibility lies not only with model but also with developers and reviewers who deploy code into production.

Blockchain security firm SlowMist pointed to vulnerability in oracle formula and breakdown in human oversight. Previous research identified dozens of critical flaws in applications built using popular AI coding tools.

Notably, in December 2025 Anthropic disclosed that earlier Claude version was able to independently identify and exploit smart contract vulnerabilities in simulated environment.

Breakthrough and new risks

Launch of Claude 4.6 underscores how quickly AI tools are becoming embedded in financial infrastructure development. Expanded reasoning and coding capabilities lower barriers for building complex applications.

At same time, Moonwell case highlights core risk: automation without rigorous auditing can scale errors as quickly as innovation. For DeFi sector, incident signals need for stricter review processes and layered oversight when integrating AI into development workflows.

On Feb. 5, 2026, OpenAI and Anthropic released their latest flagship coding models on same day.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.
Weekly Top Bonuses
up to $2,500
deposit bonus for all clients
CLAIM BONUS
Your capital is at risk.