Particle.news

FBI and Google Warn AI Agent Frameworks Are Speeding Up Cyberattacks

Authorities say autonomous tools like ARTEX cut the time to high-level access and require immediate changes to account and vendor controls.

Overview

  • The FBI issued a FLASH advisory on Nov. 13 saying ARTEX, an open-source agent framework, can take an intrusion from initial targeting to privileged access in under three hours.
  • Google’s Threat Intelligence Group reported this week that autonomous LLM frameworks now run continuous, machine-speed scanning and exploit development with little human input.
  • Security teams have tied ARTEX to real incidents, including a ConnectWise-related operation that automated attempts against more than 2,250 Microsoft 365 customer environments and intrusions that exposed about 67,000 South Korean financial customers.
  • Copies, forks and more than 1,600 exposed ARTEX consoles remain online despite the project’s public GitHub removal, which keeps easy-to-run attack tooling available to low-skill actors.
  • The FBI and vendors urge urgent steps — enable multifactor authentication, rotate secrets, limit admin privileges, patch vulnerable software, and monitor for automated login and data-exfiltration signals — because defenders now have only hours to detect and contain AI-accelerated intrusions.