Anthropic hacked to deliver remote access trojan
Points and comments are a snapshot, not live.
A malicious version of Anthropic's Claude Code extension delivered a remote access trojan via a supply chain attack.
Anthropic's verified account uploaded a malicious version (2.1.313) of the Claude Code VS Code extension to OpenVSX and Cursor's marketplace, impacting around 5,000 developer systems. The extension added an attacker-controlled SSH key, enabled reverse SSH tunnels, and exfiltrated host/user info. It auto-updated in Cursor. Anthropic silently replaced it with version 2.1.314 8 hours later but did not remove backdoors from infected systems or inform customers. Cursor stated the malicious code was not developed by Anthropic. Network IoCs were published.
What commenters are saying
Commenters debated responsibility: some argued Anthropic's verified account was hacked, others pointed to Cursor's marketplace or OpenVSX. Many criticized the lack of extension signing and security on OpenVSX. Some doubted article's claims, but Binance's confirmation and missing version numbers supported Anthropic compromise. Several noted the need for sandboxed development environments and offered tips for detection.