The Hacker News
Google reports hackers using AI agents to automate exploits
Google's Threat Intelligence Group (GTIG) says adversaries are increasingly using open‑source agentic frameworks and large language models (LLMs) to automate and accelerate cyber operations. In a report shared with The Hacker News, Google notes that attackers have moved beyond manual prompting to deploy autonomous multi‑agent offensive toolkits such as ARTEX, CAIRN, CyberStrikeAI and PentAGI to conduct multi‑stage intrusions at machine speed.
The use of these agentic stacks enables threat actors to chain different attack paths, scale operations and shorten the window between discovery of a vulnerability and active exploitation.
GTIG highlights that several actors, including financially motivated groups and suspected Chinese espionage entities, have experimented with these frameworks, along with tools like ZeroHollow (an inference service for exploit development and LLM routing) and Agent Zero (an open‑source agent framework with autonomous code execution and dynamic tool generation).
The findings also emphasise a shift towards automating patch analysis and PoC weaponisation via LLM‑driven pipelines, effectively compressing the timeline from patch disclosure to exploitation.
Google notes that attackers may use LLMs to rapidly weaponise patched n‑day vulnerabilities by focusing on a constrained set of sources—such as release notes, security research, and PoC code—while targeting AI infrastructure to profit from LLM credential and cloud resource theft, sold through underground proxies like APIMart and JetNight.
The report also points to the Jieyou platform, a Chinese‑language phishing platform that integrates an LLM chatbot to help configure phishing panels and templates, illustrating how PhaaS providers are using LLMs as a modular multiplier for lower‑skilled affiliates.