Cursor patched two critical prompt injection flaws

Security researchers Cato AI Labs identified two vulnerabilities in Cursor, an AI code editor, allowing zero-click command execution. Both issues tracked as CVE-2026-50548 and CVE-2026-50549 received a 9.8 severity rating before patches arrived in version 3.0 released April 2. The flaws let attackers escape the sandbox via prompt injection through Model Context Protocol servers or web search results, granting full control over developer machines.