AI assistant 'Cursor' abused for cyberattacks — German firm Teckentrup affected
Russian-speaking hackers used the AI programming assistant 'Cursor' to launch ransomware attacks on at least seven companies, including the German garage door manufacturer Teckentrup.
Russian-speaking hacker group misuses AI for ransomware attacks
Russian-speaking hackers have deployed the AI programming assistant 'Cursor' to conduct cyberattacks on at least seven companies in Europe and the USA — including the German garage door manufacturer Teckentrup. This is reported by Süddeutsche Zeitung, citing data exclusively available to news agency Reuters and reports from IT security firms Gambit Security and Cloudsek from last Thursday.
'Cursor' is based on technology from AI provider Anthropic, specifically the Claude Sonnet 4.5 model. The attacks by the ransomware group 'Aur0ra' took place between 8 April and 21 May this year. In addition to Teckentrup, victims included a Belgian chemical company and firms in the USA, Scotland, Italy and Argentina.
According to Cloudsek, the hacker group recorded a total of at least 20 victims. How many of these were attacked with the aid of AI remained unclear initially. Gambit and Cloudsek did not name the affected companies; however, Reuters was able to identify six of them after the agency independently verified parts of the chat logs.
The attackers circumvented the AI system's security measures using a simple deception: they told the software that the actions were merely a simulation. From logs on an unprotected server, it emerged that the AI agent subsequently executed hundreds of actions. "This is a test environment, so it is legal," the AI concluded in one of the logs itself.
In the attack on Teckentrup, the software recommended the use of a known malware programme and assessed the success probability as "very high". At least one of the affected companies appeared on the hackers' leak site — an indication of a failed extortion attempt.
The use of such AI tools gives cybercriminals a significant advantage because they can skip many manual steps and thus operate 30 to 50 per cent faster, explained Eyal Sela of Gambit Security. AI-supported hacking attacks are now the new normal, said Gambit strategy chief Curtis Simpson: "We will see something like this far more frequently in the future."
The incidents raise questions about the security of AI tools — all the more so because 'Cursor' was only integrated into SpaceX, the company led by Elon Musk, at the beginning of the month. Neither the affected companies nor SpaceX, the current owner of 'Cursor', initially commented on the incidents.
Source: Süddeutsche Zeitung