"These frameworks share several core capabilities: discovering and exploiting low-complexity vulnerabilities, using LLMs to generate single-use payloads on-demand, and using real-time execution feedback to dynamically adjust tactics," the Google Threat Intelligence Group (GTIG) warned in its quarterly GTIG AI Threat Tracker published on November 16.
GTIG's timeline: three to six months for fully agentic intrusions
Google's GTIG reported a "paradigm shift" in the last quarter and said it expects "sophisticated, fully autonomous agentic attacks to emerge in the next three to six months." That projection is drawn from a combination of Mandiant incident response telemetry, global threat actor tracking and live platform defenses compiled for the GTIG AI Threat Tracker.
ARTEX and open-source multi-agent frameworks on the rise
The report names ARTEX and open-source frameworks including CAIRN, CyberStrikeAI and PentAGI as increasingly popular tools among threat actors. GTIG said groups are also experimenting with "custom-built harnesses." These frameworks coordinate multiple sub-agents to plan, execute and adapt across many stages of an intrusion, amplifying both speed and scale.

This site is the portfolio.
OSINTSights runs on Cloudflare Workers, D1, R2, and Vectorize, with an AI pipeline on Hetzner ARM. Nubivance designed, built, and operates it. We do the same for clients.
See what we buildAutomation across the kill chain: reconnaissance to objective completion
GTIG described how frameworks "effectively manage a collection of sub-agents to plan, execute and adapt across multiple kill chain stages including reconnaissance and scanning, initial access, persistence and privilege escalation, internal reconnaissance and lateral movement, and completion of objectives." While the mechanics — for example, using compromised credentials and targeting unpatched systems — remain the same, GTIG warned that the velocity and scale introduced by these frameworks create "a new defensive challenge."
Current footprint: noisy attacks today, stealthier operations tomorrow
The group observed that many current operations trade stealth for velocity, producing "plenty of forensic signals, 'recognizable identifiers' and 'code-level traces,'" which defenders can use to detect and mitigate intrusions. GTIG cautioned, however, that stealth is likely to increase as capabilities evolve. The report adds that the "low barrier to entry, ease of integrating new exploits and massive scaling potential" will make these frameworks persistently attractive to both cybercriminals and nation-state operators.
How SOC teams, the FBI, and South Korean financial institutions are responding
- Security operations and SOC teams: GTIG said its own SecOps team has "deployed specialized behavioral and volumetric rulepacks to detect use of ARTEX." Exabeam's Sally Vincent, senior threat research engineer, recommended improved behavioral monitoring to "flag unfamiliar tool use or unusual data access" and to "identify changes in activity volume" that can expose attacks even when legitimate credentials or tools are abused.
- Incident responders and vendors: Mandiant has released a hardening guide and a defensive framework aimed at countering the "speed and scale advantage" of AI-powered attackers, according to GTIG's compilation of Mandiant incident response telemetry.
- Law enforcement and affected institutions: GTIG noted that the FBI issued a warning last week about the growing use of ARTEX in attacks, and that South Korean financial institutions have been cited among recent victims.
The GTIG tracker highlights a shifting calculus: today’s attacks often shout their presence through speed and detectable artifacts; tomorrow's aim appears to be quiet, autonomous campaigns that adapt in real time. Defenders have begun to respond—deploying behavioral rulepacks, publishing hardening guides and sharing warnings—but the report frames this as a race against rapid automation and scaling.
GTIG’s concrete examples—named frameworks, tactics that automate each kill-chain stage, vendor and law-enforcement actions—make the timeline explicit. The central question the tracker leaves on the table is operational: as multi-agent frameworks mature into fully autonomous intrusions over the next three to six months, will the combination of behavioral detection, platform defenses and hardening guidance be enough to blunt that advantage?




