Russian-speaking hackers used SpaceX AI tool Cursor to breach multiple companies

Russian-speaking hackers from a ransomware group called Aur0ra used the AI coding assistant Cursor, a tool recently acquired by Elon Musk's SpaceX, to breach at least seven companies globally. According to reports from cybersecurity firms Gambit Security and CloudSek, the attackers tricked the AI agent into performing hundreds of malicious operations, including stealing login credentials and recommending malware, by convincing the software that the attacks were part of a simulation. The AI agent, which was powered by Anthropic's Claude Sonnet 4.5 model, reportedly refused some requests it deemed illegal but was bypassed when the hackers claimed the environment was for testing.

Chat logs from April 8 to May 21 identified victims in the United States, Germany, Belgium, Scotland, Italy, and Argentina. Specific companies identified by Reuters include the German garage door manufacturer Teckentrup, the Belgian cleaning products maker Christeyns, and the Scotland-based Helideck Certification Agency. While Gambit Security estimated that the AI tool likely increased the hackers' speed by 30 to 50 percent, CloudSek reported that the Aur0ra group may have claimed as many as 20 victims in total, though it is unclear how many of those involved the AI tool.

The breach was discovered after the hackers inadvertently left one of their servers unsecured. Neither SpaceX, Cursor, nor Anthropic responded to requests for comment. Gambit Security's chief strategy officer, Curtis Simpson, described the situation as a cat and mouse game, noting that AI-supported attacks are becoming a new normality.