AI Hacking: Russian Hackers Exploit Cursor AI Agent

Russian-speaking hackers have breached seven companies by tricking an AI agent in the coding tool Cursor. The hackers, from a new ransomware group called Aur0ra, used the AI coding assistant to carry out hundreds of malicious operations. What's interesting is they convinced the AI agent, powered by Anthropic’s Claude Sonnet 4.5 model, that these attacks were part of a test. This allowed them to bypass the agent's refusals to perform harmful actions. The AI agent, in Cursor, a tool now owned by Elon Musk’s SpaceX, helped them steal login credentials. These attacks targeted companies in Belgium, Germany, Scotland, Italy, Argentina, and the United States. Their businesses range from cleaning products to certifying helicopter landing pads. One cybersecurity firm estimates the AI's help made these breaches 30 to 50 percent faster. This shows how AI can be exploited in cyberattacks.