Threat actors are using the recent Claude Code leak as a social engineering lure to distribute malware via GitHub repositories. They create trojanized versions of the leaked Claude source code that include malicious payloads such as Vidar stealer v18.7 and the GhostSocks trojan. This campaign demonstrates rapid opportunistic exploitation of a high-profile code leak, with compromised GitHub repositories serving as delivery mechanisms. Command and control infrastructure has been identified across multiple IP addresses and domains. The threat does not stem from a software vulnerability but from malicious repackaging and social engineering. Multiple indicators including IPs, domains, URLs, and file hashes have been identified to help detect this activity.

The impact involves potential compromise of systems that download and execute the trojanized Claude code from malicious GitHub repositories. The delivered malware includes information stealers (Vidar stealer) and trojans (GhostSocks), which can lead to data theft, unauthorized access, and further compromise. The threat leverages social engineering tied to a high-profile AI code leak to increase victim trust and likelihood of execution. There is no direct vulnerability in the original Claude code; the risk arises from malicious actors repackaging and distributing malware under the guise of leaked code.

There is no official patch or fix because this threat involves malicious use of leaked code and social engineering rather than a software vulnerability. Organizations should implement Zero Trust architecture principles to reduce risk from shadow AI instances and trojanized agents. Avoid downloading or executing code from untrusted or unofficial GitHub repositories, especially those claiming to contain leaked Claude source code. Monitor for and block known indicators of compromise such as the listed IP addresses, domains, URLs, and file hashes. User education on social engineering risks related to high-profile leaks is also recommended.