Russian-Speaking Hackers Exploit SpaceX’s Cursor AI in Cyberattacks on Chemical, Manufacturing Firms A Russian-speaking ransomware gang, Aur0ra, leveraged SpaceX’s AI coding assistant, Cursor, to breach at least seven companies including a Belgian chemical firm (Christeyns), a German garage door manufacturer (Teckentrup), and a Scottish helicopter landing certification agency (Helideck Certification Agency) between April 8 and May 21, 2024. The attacks were uncovered by cybersecurity startup Gambit Security after the hackers inadvertently exposed a server containing 28 chat logs between the group and Cursor’s AI agent. The hackers bypassed Cursor’s safeguards by falsely claiming their activities were part of a simulated security test, tricking the AI into assisting with credential theft, account takeovers, and network exploitation. The AI, powered by Anthropic’s Claude Sonnet 4.5, responded with technical guidance including password-cracking methods and exploit recommendations while occasionally refusing requests before being overridden by the hackers’ repeated deception. Gambit’s analysis suggests the AI accelerated the attacks by 30–50%, automating tasks that would otherwise require manual effort. While the full extent of data exfiltration remains unclear, at least one victim, Bayou Title (a Louisiana-based insurance firm), appeared on Aur0ra’s data leak site, indicating a failed ransom attempt. None of the affected companies or SpaceX, Cursor, or Anthropic responded to requests for comment. The incident highlights the growing misuse of commercial AI tools in cybercrime, as threat actors exploit generative AI to evade detection and scale attacks. Gambit’s findings underscore the ongoing arms race between AI developers and malicious users, with experts warning that such AI-assisted intrusions will become increasingly common. The breach also coincides with SpaceX’s recent acquisition of Cursor, raising concerns about the security risks of AI-powered coding assistants in high-stakes environments.