Hackers Use Fake Claude Code Leaks to Spread Malware

Exploiting the Claude Code Hype

Security researchers at Zscaler ThreatLabz recently uncovered a targeted campaign using Anthropic’s new developer tool as bait. Threat actors are distributing malware by promising early access to leaked versions of Claude Code. This activity highlights how quickly attackers pivot to exploit high-interest software releases in the engineering community.

The campaign primarily targets developers on platforms like GitHub and social media. Attackers create repositories that appear to host the source code or binary for the command-line tool. Instead of functional software, these files contain scripts designed to compromise local environments and steal sensitive credentials.

Tactics and Technical Execution

The attackers use several methods to increase the perceived legitimacy of their malicious files. By mimicking the official documentation and branding of Anthropic, they trick users into running installation scripts. These scripts often perform the following actions:

• Exfiltrating browser cookies and saved passwords from the victim's machine.
• Scanning for environment variables and API keys stored in local configuration files.
• Establishing a persistent backdoor for future remote access.
• Deploying secondary payloads like info-stealers or ransomware.

The scripts are frequently obfuscated to bypass basic antivirus scans. Developers often grant high-level permissions to command-line tools, making this specific attack vector particularly dangerous for corporate security.

Risks to Development Environments

This incident underscores the danger of side-loading tools outside of official package managers like NPM or PyPI. Because Claude Code is a productivity tool, developers may feel pressured to bypass security protocols to gain a competitive edge. Compromising a developer's machine can lead to broader supply chain attacks if the intruder gains access to internal company repositories.

Security teams should enforce strict policies against downloading software from unverified third-party sources. Using sandboxed environments or virtual machines for testing new tools can mitigate the risk of host infection. Organizations must also monitor for unauthorized outbound traffic to known command-and-control servers identified by threat intelligence reports.

Monitor official Anthropic communication channels for the verified release and checksums of the Claude Code utility.