Israeli researchers found a critical flaw in Grok – and were completely ignored

Researchers from Profero uncovered critical security vulnerabilities in xAI's Grok bot, allowing full server takeover. Despite repeated reports, xAI and Cursor ignored the findings and failed to fix the seven vulnerabilities for over two months.

Researchers from the Israeli cybersecurity company Profero, led by founder Omri Segev Moyal, have uncovered a series of severe security vulnerabilities in the core of the AI agent Grok bot, developed by Elon Musk's xAI. According to a report published Monday, the agent's infrastructure, operated by SpaceX-owned Cursor, suffers from design flaws that allow hackers to gain full administrative privileges on the virtual machine and control the user account from any computer. The researchers emphasize that the flaw functions as a combination of a Trojan horse and a built-in rootkit, without the need for external malware. Among the findings: a default password under the word 'local', storage of login data in plain text, and bypass of encryption mechanisms. The researchers contacted xAI and Cursor as early as late August, but xAI ignored all seven reports, and Cursor only provided an automated response. Over two months later, both companies released more than 240 software versions without fixing any of the seven critical vulnerabilities.