Google’s Threat Intelligence Group has reported the first known case of an AI framework fully automating vulnerability discovery and exploitation without human involvement. In a report titled “Agentic Threat Activity Crosses a Significant Threshold,” researchers described a multi-agent system called exploit_agent found on an exposed staging server. The framework connects to multiple large language models and continuously compares the code of popular web applications before and after security patches. Once it identifies a potential flaw, it automatically drafts and executes an exploit. The system also includes scripts for post-compromise activities such as internal reconnaissance, brute-forcing, and installing or removing webshells. Evidence pointed to a preference for French media, entertainment, and education websites, with a Russian-language comment in the configuration file. The findings underscore a shrinking window between patch release and active exploitation.
Google’s Threat Intelligence Group has reported the first known case of an AI framework fully automating vulnerability discovery and exploitation without human involvement. In a report titled “Agentic Threat Activity Crosses a Significant Threshold,” researchers described a multi-agent system called exploit_agent found on an exposed staging server. The framework connects to multiple large language models and continuously compares the code of popular web applications before and after security patches. Once it identifies a potential flaw, it automatically drafts and executes an exploit. The system also includes scripts for post-compromise activities such as internal reconnaissance, brute-forcing, and installing or removing webshells. Evidence pointed to a preference for French media, entertainment, and education websites, with a Russian-language comment in the configuration file. The findings underscore a shrinking window between patch release and active exploitation.