Google Warns Autonomous AI Attacks Have Arrived
Google has warned that the era of autonomous cyberattacks has arrived, after discovering an AI-powered framework capable of identifying software vulnerabilities, writing exploits, and executing attacks without human intervention. The framework, called exploit_agent, was found on an exposed staging server and uses multiple large language models to compare code before and after security patches, reverse-engineering fixes to target unpatched systems. Google also found exfiltrated victim data and post-compromise scripts on the server, with the operator showing interest in French media and education websites. Google notes these tools currently trade stealth for speed, generating detectable forensic traces, and says defenders will need AI agents to keep pace. The discovery follows earlier incidents of AI-assisted attacks, including a September 2026 campaign that compromised hundreds of PaperCut servers.