AI news · Claude Code · Cursor marketplace
Cursor pulls a fake Claude Code update that planted SSH keys
A modified copy of Anthropic’s coding extension sat in Cursor’s marketplace for about 11 hours on 29 October, reaching at most 5,000 users, Cursor says.
For about 11 hours on 29 October 2026, the extension marketplace inside Cursor, a popular AI code editor, handed out a doctored copy of Claude Code, Anthropic’s coding assistant, that tried to open a back door on every computer that installed it. Versions 2.1.312 and 2.1.313 for Windows and macOS were live from about 2:27am to 1:35pm Pacific time, Cursor’s notice says, and Cursor put the number of affected users at 5,000 at most, The Hacker News reports.
The package slipped an attacker’s SSH key (the credential that lets someone log in to a machine remotely) onto each computer, Cursor says, together with start-up scripts that retried a hidden connection to the attacker, called a tunnel, about every 60 seconds and sent the machine’s details back. Bugs in the code and missing administrator rights stopped that connection on most machines, but the key and the start-up scripts still landed, and Cursor warns that uninstalling the extension may not be enough because the scripts keep running on their own. Cursor shut, rebuilt and re-keyed its marketplace.
Microsoft’s threat intelligence team says the doctored versions never appeared on its own VS Code Marketplace or on npm, and that Anthropic told Microsoft its systems were not breached.
Claude Code comes with Claude Pro at $20 a month on our 1 November check, and the setup page for Meta’s ads MCP connector (a standard that plugs AI assistants into other software), which we read on 26 September, lists Claude Code among four agents that can work on Meta ad accounts.
The key and the start-up scripts outlive the extension, so anyone who installed or updated Claude Code through Cursor on 29 October has two jobs left after uninstalling it: run Cursor’s check for the planted key, labelled svc-telemetry, and its clean-up script.
Sources
- Cursor: Security notice: malicious version of Anthropic.claude-code (2.1.312 and 2.1.313) identified in the Cursor VS Code Extension Marketplace. Checked
- Microsoft Security Blog: Inside a malicious Claude Code-branded extension and its persistent SSH backdoor. Checked
- The Hacker News: Official Claude Code Extension on Cursor Marketplace Trojanized with Persistent SSH Backdoor. Checked
More AI news from
- Checkout appears in a quarter of AI agent visits, Clarity finds
- Google trials one Sponsored heading for shopping ads in AI Mode
- ChatGPT share links could hijack accounts, Parameter says
- Google Search panels now offer Buy now through Universal Cart
- Bluefish turns AI answer gaps into ChatGPT and Google ad briefs
- Gemini stops tagging cited links with utm_source=gemini
- Deepgram buys Pipecat maker Daily and keeps the framework open
- Koah says gross sales on its AI chat ad network rose fourfold
- Resemble AI says bot tools missed most disguised AI agents