Google’s latest AI threat tracker report revealed that bad actors are now using autonomous large language model (LLM) frameworks to continuously scan software updates, identify patched vulnerabilities and develop attacks targeting those flaws, all without human involvement.
The report said threat actors have moved beyond manual prompting to deploying autonomous, multi-agent offensive frameworks such as ARTEX, CAIRN, CyberStrikeAI and PentAGI to carry out multistage intrusions at greater speed.
The misuse of open-source, agentic penetration-testing (pentesting) frameworks is emerging as a way to “automate discrete phases and specific operational tasks during live intrusions,” Google Threat Intelligence Group (GTIG) said.
Google said the adoption represents a paradigm shift in which cyberattacks occur at machine speed. This allows threat actors to scale operations significantly by combining different attack paths, even as the underlying methods remain largely unchanged.
READ: Microsoft struggles to restore Outlook, Teams and Copilot after global outage hits 74 countries (November 13, 2026)
“These observations additionally extend to the use of other tools and services, such as ZeroHollow, an inference service used for exploit development and LLM routing, and Agent Zero, an open source agent framework which features autonomous code execution and dynamic tool generation,” Google said.
The report identified financially motivated threat actors and suspected espionage groups aligned with China among those using these tools.
Another area of misuse involves automating labor-intensive tasks, such as vulnerability analysis and proof-of-concept (PoC) weaponization, through LLM-driven pipelines. This allows threat actors to shorten the time between the disclosure of a patch and the active exploitation of the vulnerability.
“Threat actors may find it more efficient to use LLMs to rapidly weaponize patched n-day vulnerabilities, as they can focus research on a finite set of source materials: comparing source code versions (including from pre-release or preview channels), and examining release notes, security research, and POC code related to specific flaws,” Google added.
READ: Bill Gates warns AI without safeguards could fuel cyberattacks, calls for government oversight (October 12, 2026)
Threat actors are also targeting AI infrastructure to generate illicit profits through LLMjacking attacks. These involve stealing LLM credentials and cloud resources from compromised enterprise environments and reselling access through underground proxy services such as APIMart and JetNight at discounted rates. These services are also known as LLM mixers.
According to the findings, the emergence of open-source frameworks has further lowered the barrier to entry. These tools make it easier for bad actors to integrate new exploits, scale attacks autonomously and mount convincing phishing campaigns, compressing the time needed to carry out attacks.
John Hultquist, chief analyst at Google Threat Intelligence Group, said attackers were largely not “using new groundbreaking methods to break into systems.”
“They are still going after known, unfixed issues, just with much greater speed and scale. Instead of relying on manual effort to keep up, we will have to embrace agentic defense. Or we risk falling behind,” he said.


