Cybercriminals are taking advantage of the recent Claude Code source code disclosure by spreading the Vidar data-stealing malware through deceptive GitHub repositories.

Claude Code is Anthropic’s AI tool that operates within the terminal, enabling code execution and acting as an independent agent. It can interact with systems, manage LLM API calls, integrate with MCP, and remember information.

On March 31, Anthropic mistakenly released the complete source code of the client-side component of the tool. This happened because a 59.8 MB JavaScript source map was inadvertently included in the npm package.

The leaked code consisted of 513,000 lines of easy-to-read TypeScript code across 1,906 files. It exposed how the agent works, its access rights, how it executes tasks, hidden functions, build information, and internal security details.

The exposed code was quickly obtained by many users and shared on GitHub, where it was duplicated thousands of times.

According to a report by cloud security firm Zscaler, the leak has provided threat actors with an opportunity to distribute the Vidar infostealer to users searching for the Claude Code leak.

Researchers discovered a fake leak posted on a malicious GitHub repository by user “idbzoomh,” which falsely claimed to have “unlocked enterprise features” and no usage restrictions.

To attract more users, the repository is optimized for search engines, causing it to appear prominently in Google Search results for terms like “leaked Claude Code.”

Researchers say that unsuspecting users download a 7-Zip archive containing a Rust-based executable named ClaudeCode_x64.exe. When executed, it deploys Vidar, a common type of information-stealing malware, along with the GhostSocks network proxy tool.

Zscaler noted that the malicious archive is frequently updated, suggesting that other malicious programs could be added in the future.

The researchers also found a second GitHub repository with the same code, but instead of a functional download link, it displays a non-working “Download ZIP” button. Zscaler believes the same threat actor operates this repository and is likely testing different distribution techniques.

Despite security measures, GitHub has frequently been used to spread malware disguised in various ways.

In campaigns during late 2025, less experienced researchers or cybercriminals were targeted with repositories purporting to contain proof-of-concept (PoC) exploits for newly discovered vulnerabilities.

Historically, attackers have been quick to take advantage of high-profile events, hoping to compromise systems opportunistically.