Developers using Claude AI through Visual Studio Code have encountered a significant security threat following the discovery of a malicious version of this popular coding assistant. Cybersecurity experts revealed the presence of a concealed backdoor in version 2.1.313 of the Claude Code extension.
Security Breach Detected on Open VSX
In a report dated October 30, researchers from OpenSourceMalware announced the detection of this vulnerable version within the Open VSX extension registry. Known for deploying tools to code editors, Open VSX is a widely used extension hosting platform among developers.
The researchers emphasized that the malicious version closely mimics the legitimate tool in both appearance and basic functionality, complicating the identification of this security threat.
“OpenSourceMalware researchers confirmed the malicious code in the 2.1.313 version examined on Open VSX but were unable to ascertain how it was integrated into the system,” they stated.
Upon installation, the injected code attempts to establish a remote connection to a server controlled by attackers. If successful, this connection could grant unauthorized access to the affected computer.
Persistent Threat Amplifies the Risk
The researchers noted that the malware is engineered to remain active even after the computer restarts. Thus, simply closing VS Code or ceasing to use Claude does not necessarily mitigate the threat.
Due to this, merely uninstalling the extension is deemed insufficient. Developers who have installed version 2.1.313 via Open VSX are urged to consider their systems potentially compromised and initiate an immediate security audit.
Potential Threats for Crypto Developers
No evidence suggests that the attack was specifically targeted at cryptocurrency users. However, developers working on blockchain projects may have private keys, wallet login credentials, exchange API keys, and access information to crypto infrastructures on their machines, elevating their risk level.
Should attackers gain remote access, there is a possibility for these details to be exfiltrated and utilized in unauthorized transactions involving wallets or projects.
The researchers did not detect blockchain-specific functions within the malicious code nor reported any confirmed cryptocurrency theft cases.
Nevertheless, the current findings do not substantiate allegations that Anthropic’s official VS Code Marketplace distribution or its broader distribution networks have been compromised. The incident appears confined to the specific version on Open VSX at this stage.
