Google’s Threat Intelligence Group (GTIG) has issued a warning that fully autonomous, agentic cyberattacks are likely to emerge within the next three to six months. This projection, detailed in the company's quarterly AI Threat Tracker released on November 16, 2026, follows a significant shift in how threat actors are utilizing automated penetration testing frameworks. By synthesizing incident response telemetry from Mandiant and global threat actor tracking, Google identified a trend toward the use of sophisticated tools that can plan, execute, and adapt across the entire cyber kill chain.
The frameworks currently gaining traction among both cybercriminals and nation-state actors include ARTEX, as well as open-source, multi-agent systems such as CAIRN, CyberStrikeAI, and PentAGI. According to Google, these tools allow attackers to manage a collection of sub-agents that perform reconnaissance, scanning, initial access, persistence, privilege escalation, and lateral movement. Some groups are also reportedly developing custom-built harnesses to further refine these capabilities.
These frameworks share several core technical features that distinguish them from traditional automated scripts. They are capable of identifying and exploiting low-complexity vulnerabilities, using large language models to generate single-use payloads on demand, and employing real-time execution feedback to dynamically adjust tactics during an intrusion. While the underlying mechanics of these attacks—such as the use of compromised credentials and the targeting of unpatched systems—remain consistent with historical methods, the speed and scale at which these frameworks operate present a new challenge for defensive teams.
Google confirmed that it has observed a notable increase in the use of ARTEX over the past quarter. The FBI has corroborated these findings, recently issuing an alert regarding the growing abuse of the ARTEX framework, specifically noting its use in campaigns targeting South Korean financial institutions.
Despite the advanced nature of these tools, Google noted that current attacks often prioritize velocity over stealth. Consequently, these operations frequently generate significant forensic signals, recognizable identifiers, and code-level traces that allow defenders to detect and mitigate the activity. However, the report cautions that as these frameworks evolve, attackers will likely prioritize stealth, making detection significantly more difficult in the future.
To counter the current wave of AI-powered threats, Google’s security operations team has deployed specialized behavioral and volumetric rulepacks specifically designed to detect the use of ARTEX. Mandiant has also published a hardening guide and a defensive framework intended to help organizations mitigate the speed advantage held by these automated systems.
Industry analysis suggests that security operations centers can bolster their defenses by focusing on behavioral monitoring. By flagging unfamiliar tool usage, unusual data access patterns, and sudden changes in activity volume, organizations may be able to identify and block these threats even when attackers utilize legitimate tools or valid credentials to carry out their objectives.
Google emphasizes that the low barrier to entry, the ease of integrating new exploits, and the massive scaling potential of these agentic frameworks will likely make them a permanent fixture in the threat landscape. As these capabilities continue to mature, the company expects these operations to move further along the autonomous scale, ultimately resulting in fully agentic intrusions that require little to no human intervention.





