Ethical Hacking News
A recent GitHub leak has exposed sensitive information about an AI agent's inner workings, allowing malicious actors to deliver malicious payloads to unsuspecting users. Learn more about the Claude Code leak and its implications for security.
The Claude Code source code leak exposed AI agent's inner workings to malicious actors. The leaked code contained sensitive information about the Claude Code agent, including orchestration logic and permissions. A malicious repository was created using the leaked code, containing a Rust-based executable that deployed the Vidar infostealer and GhostSocks network traffic proxying tool. Another identical GitHub repository was discovered, likely used as a testing ground for delivery strategies. The leak highlights the vulnerabilities of open-source software and the importance of secure coding practices.
In a shocking turn of events, the recent Claude Code source code leak has exposed an AI agent's inner workings to malicious actors. The leak, which occurred on March 31, 2026, was accidentally published by Anthropic in their npm package, revealing over 513,000 lines of unobfuscated TypeScript across 1,906 files.
The leaked source code contained sensitive information about the Claude Code agent, including its orchestration logic, permissions, and execution systems. This information has been used by threat actors to deliver malicious payloads to unsuspecting users. One such payload is the Vidar infostealer, a commodity information stealer that can be deployed through the compromised GitHub repository.
The malicious repository, created by user "idbzoomh", advertises itself as having "unlocked enterprise features" and no usage restrictions. However, upon closer inspection, researchers from cloud security company Zscaler discovered that the repository is optimized for search engines and appears among the first results on Google Search for queries related to the leaked Claude Code.
The malicious archive downloaded by users contains a Rust-based executable named ClaudeCode_x64.exe, which deploys Vidar along with the GhostSocks network traffic proxying tool. The researchers found that the malicious archive is updated frequently, indicating that other payloads may be added in future iterations.
Another identical GitHub repository was discovered, operated by the same threat actor who likely experiments with delivery strategies. This second repository shows a 'Download ZIP' button that wasn't functional at the time of analysis, suggesting that it may have been used as a testing ground for malicious payloads.
The Claude Code leak has highlighted the vulnerabilities of open-source software and the importance of secure coding practices. As we move forward in an increasingly digital world, it is crucial to remain vigilant and proactive in protecting ourselves against such threats.
Related Information:
https://www.ethicalhackingnews.com/articles/Claude-Code-Leak-Exposes-AI-Agents-Secrets-to-Malicious-Actors-ehn.shtml
https://www.bleepingcomputer.com/news/security/claude-code-leak-used-to-push-infostealer-malware-on-github/
https://www.theregister.com/2026/04/02/trojanized_claude_code_leak_github/
Published: Thu Apr 2 16:15:05 2026 by llama3.2 3B Q4_K_M
