Google expects sophisticated, fully autonomous agentic attacks to emerge within the next three to six months after observing a significant paradigm shift in how threat actors use penetration-testing frameworks.
The Google Threat Intelligence Group shared these findings in its quarterly AI Threat Tracker published on November 16, Infosecurity Magazine reported.
The report relies on Mandiant incident response telemetry, global threat tracking, and live platform defenses. Analysts warn that tools like ARTEX and open-source multi-agent frameworks such as CAIRN, CyberStrikeAI, and PentAGI are growing rapidly in popularity among malicious groups.
“While the mechanics of the intrusions remain the same, such as using compromised credentials and targeting unpatched systems, the speed and scale at which frameworks can exploit these known weaknesses introduces a new defensive challenge,” the report explained.
These frameworks manage collections of sub-agents to plan, execute, and adapt across multiple kill chain stages. These stages include reconnaissance, initial access, privilege escalation, and lateral movement.
Russian-Speaking Hackers Run First Automated AI Intrusion
Google identified what analysts believe is the first end-to-end cyberattack run entirely by artificial intelligence without human intervention, according to CyberScoop. Researchers discovered a system dubbed “exploit_agent” on an exposed staging server containing stolen victim data.
The tool chains together multiple AI models to continuously compare software code before and after vendors release patches. When a fix appears, the system infers what the change reveals, drafts an exploit, and executes it automatically.
Investigators found Russian-language comments in a configuration file instructing users to enter API keys for auto-exploitation. Victim data on the server pointed to targeting of French media, entertainment, and education websites.

Attackers are also exploiting the velocity of patched code. After vendors fixed two PHP vulnerabilities on October 21, hundreds of purported exploits appeared within days, most of them nonfunctional and likely generated by AI.
Defending Against Machine-Speed Exploits
Threat actors currently trade stealth for velocity, generating recognizable forensic signals and code-level traces that help defenders detect intrusions. However, security analysts warn that stealth will increase as the technology matures.
“It can also identify changes in activity volume. These changes can help expose an attack even when the attacker uses legitimate tools and valid credentials,” said Sally Vincent, senior threat research engineer at Exabeam, told Infosecurity Magazine.
Google has deployed specialized behavioral and volumetric rulepacks to detect the use of ARTEX across its security operations. Mandiant also released a hardening guide designed to help organizations counter the speed of AI-powered attackers.
To address machine-speed threats, Google introduced its Google AI Threat Defense system. “We're helping defenders counter automated threats with Google AI Threat Defense, which combines Gemini and other frontier models to find and fix flaws before attackers can exploit them,” Google said.
AI frameworks and exploit systems used in attacks
What specific AI frameworks are threat actors using in live attacks?
Threat actors are actively using penetration-testing frameworks like ARTEX alongside open-source multi-agent systems such as CAIRN, CyberStrikeAI, and PentAGI.

How does the newly discovered exploit_agent system operate?
The exploit_agent system uses AI models to continuously compare software code before and after a patch is published, draft an exploit based on the fix, and run it without human involvement.
Which sectors have been targeted by these automated AI campaigns?
Victim data recovered from staging servers pointed to targeting of French media, entertainment, and education websites, alongside broader activity affecting US aerospace and South Korean financial institutions.
What defense mechanisms are security teams deploying against these threats?
Google’s security operations team has deployed specialized behavioral rulepacks to detect ARTEX usage, while Mandiant released a hardening guide and Google AI Threat Defense integrates Gemini models to patch flaws automatically.