Malicious Claude Code extension hits Cursor marketplace

Attackers are increasingly targeting AI coding tools and extensions, such as an incident involving Anthropic's Claude Code extension on Cursor's marketplace and campaigns tracked by security firm Wiz against tools like Trivy and LiteLLM. Because modern AI agents and developer assistants possess broad permissions, shell execution access, and cloud credentials, a single compromised tool can provide attackers with deep access to sensitive infrastructure. To mitigate these risks, organizations must apply zero trust to dependencies, own their software supply chains, control updates, limit blast radii, maintain incident response readiness, and monitor development environments. Open article →