Tens of thousands of people eagerly downloaded the leaked Claude Code source code this week, and some of those downloads came with a side of credential-stealing malware.

A malicious GitHub repository published by idbzoomh uses the Claude Code exposure as a lure to trick people into downloading malware, including Vidar, an infostealer that snarfs account credentials, credit card data, and browser history; and GhostSocks, which is used to proxy network traffic. 

Zscaler's ThreatLabz researchers came across the repo while monitoring GitHub for threats, and said it's disguised as a leaked TypeScript source code for Anthropic's Claude Code CLI. 

"The README file even claims the code was exposed through a .map file in the npm package and then rebuilt into a working fork with 'unlocked' enterprise features and no message limits," the security sleuths said in a Thursday blog.

They added that the GitHub repository link appeared near the top of Google results for searches like "leaked Claude Code." While that was no longer the case at The Register's time of publication, at least two of the developer's trojanized Claude Code source leak repos remained on GitHub, and one of them had 793 forks and 564 stars.

• Anthropic goes nude, exposes Claude Code source by accident
• Claude Code source leak reveals how much info Anthropic can hoover up about you and your system
• Malware-laced OpenClaw installers get Bing AI search boost
• AI agents are 'gullible' and easy to turn into your minions

The malicious .7z archive in the repository's releases section is named Claude Code - Leaked Source Code, and it includes a Rust-based dropper named ClaudeCode_x64.exe.

Once it's executed, the malware drops Vidar v18.7 and GhostSocks onto users' machines, and then the Vidar stealer gets to work collecting sensitive data while GhostSocks turns infected devices into proxy infrastructure that criminals can use to mask their true online location and carry out additional activity through compromised computers.

In March, security shop Huntress warned about a similar malware campaign using OpenClaw, the already risky AI agent platform, as a GitHub lure to deliver the same two payloads.

Both of these illustrate how quickly criminals move to take a buzzy new product or news event (like OpenClaw and the Claude Code leak) and then abuse it for online scams and financial gain. "That kind of rapid movement increases the chance of opportunistic compromise, especially through trojanized repositories," the Zscaler team wrote.

The blog also includes a list of indicators of compromise, including the GitHub repositories with the trojanized Claude Code leak and malware hashes to help defenders in their threat-hunting efforts, so be sure to check that out - and, as always, be careful what you download. ®

More about

• AI
• Claude
• Cybercrime