Summary
Google said cybercriminals are using AI systems to identify software weaknesses, create malicious tools and attempt attacks with less human direction. The company said the tools can speed up attacks on unpatched systems, while some still make mistakes and leave traces security teams can detect.Key Facts
- Google researchers found an AI-powered system called exploit_agent designed to study software updates, identify newly fixed weaknesses and target systems that had not yet been updated.
- The system could attempt to exploit vulnerabilities without a hacker directing each step.
- Researchers found stolen victim data and tools for searching compromised networks, breaking into accounts and installing malicious software on a server hosting exploit_agent.
- Google observed another framework, CAIRN, in intrusion activity involving organizations that included U.S. aerospace entities.
- A separate tool based on PentAGI scanned organizations in Asia for weaknesses in their computer systems.
- Google said attackers used access to paid AI services and company cloud accounts to run AI systems, with some victims facing more than $100,000 in fraudulent charges in one day.
- Google said autonomous hacking tools can leave recognizable traces, and some AI-generated attacks fail to work.