Cursor

Researchers demonstrate how malicious instructions hidden in PNG images can hijack AI coding assistants to exfiltrate credentials. Claude Code resisted the attack while Cursor and others complied.

Wiz researchers discover six major AI coding assistants vulnerable to symlink attacks. Malicious repos can trick Claude Code, Cursor, and Amazon Q into writing files outside designated workspaces.

Two CVSS 9.8 vulnerabilities in the popular AI code editor allow zero-click attacks where malicious instructions in external data sources execute arbitrary commands on developer machines.