Fast Facts
- Cursor’s Claude Code extension was hijacked for 11 hours on Oct. 29; users who installed version 2.1.313 may have been exposed.
- The malicious extension stole system details and attempted to install a persistent SSH backdoor.
- Cursor said up to 5,000 users may be affected; it took the marketplace offline, rotated keys, and removed attacker infrastructure.
- Separately, researchers found a Cursor browser flaw that could let AI agents run arbitrary code; Cursor says it patched the issue.
Malicious Claude Code Update Reached Cursor Users
A compromised version of Anthropic’s Claude Code extension appeared in Cursor’s marketplace for 11 hours on Oct. 29. Developers who downloaded or updated the extension during that window received version 2.1.313, which added malicious code to the legitimate software.
The code collected system details, including usernames, hostnames, working folders and public IP addresses. It also tried to create a backdoor that could give attackers lasting access. However, that step required administrative privileges, which the extension usually did not have.
Cursor said up to 5,000 users may have received the affected version. The company planned to contact them directly. It also rotated security keys, checked other extensions, rebuilt its marketplace and worked to remove the attacker’s infrastructure. Cursor said Anthropic did not develop the malicious version. The company has not explained how the attacker gained access.
Security Must Keep Pace With AI Coding Tools
Coding extensions can access files, run programs and connect to the internet. Therefore, a compromised extension can pose greater risks than many browser add-ons. Security experts say attackers sometimes steal the credentials that let publishers release updates. Stronger protections, such as two-factor authentication, can help reduce that risk.
Separately, researchers reported a flaw in Cursor’s built-in browser that could let a malicious website prompt an AI agent to run harmful code. The company patched that issue on Oct. 1, according to the report. The browser flaw and the extension attack were unrelated.
AI coding tools can help developers work faster and focus on creative problems. Yet this incident shows why teams need to verify updates and protect publishing accounts. As these tools take on more tasks, their practical benefits will depend on safeguards that keep pace.
Expand Your Tech Knowledge
Learn how the Internet of Things (IoT) is transforming everyday life.
Stay inspired by the vast knowledge available on Wikipedia.
CyberRisk-V1
