🔍 Read the full analysis: AI Attacks Without Human Involvement: What Google Can And Can’t Show on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Google’s November 16 threat report describes an exposed server running exploit_agent, which it says analyzed software patches, developed exploits and executed them without human approval at each step. The finding documents one observed case; Google’s predictions that attacks will become broadly near-autonomous in coming months are forecasts, not evidence that such operations are already routine.
Google’s Threat Intelligence Group says it found an exposed server running software that analyzed security updates, developed exploits and launched them automatically, without a person approving each step. The finding, published November 16 with Mandiant, is evidence of one observed automated exploitation framework—not proof that fully autonomous cyberattacks are already common.
In its report, “GTIG AI Threat Tracker: Agentic Threat Activity Crosses a Significant Threshold,” Google describes a multi-agent system it calls exploit_agent. The server linked software before and after updates to infer which vulnerability a patch addressed, then drafted and executed an attack, according to GTIG. The framework connected to multiple commercial and locally hosted large language models.
Google says the server also held victim data and scripts for later activity, including network reconnaissance and web backdoors. Data on the server suggested its operator targeted French websites in media, entertainment and education. Google did not identify the operator or say how many sites were successfully compromised. A figure in the report shows the framework examining access checks in a WordPress plugin; the report does not establish that this examination resulted in a successful attack.
GTIG said the system was the first instance it had identified of attackers using LLM frameworks to fully automate vulnerability discovery and exploitation without a human in the loop. That wording describes Google’s reported observation, not an independently established first across all cyber operations. The report also names frameworks including ARTEX, CAIRN, CyberStrikeAI and PentAGI in its discussion of AI-enabled activity. Google says it observed CAIRN used against targets that included the U.S. aerospace sector and a customized PentAGI platform used against organizations in Asia.
Google threat report · Nov 16, 2026
How far AI attacks have come, by Google’s count
Google sorts attacker use of AI into seven stages. Four are history, one is now, two are a forecast.
Google’s seven stages
Observed · end of 2025 – Sep 2026
- 1Prompting a chatbothuman-driven
- 2AI built into malwareautomated runtime call
- 3AI automation in operationsAI in the loop, humans in charge
- 4Experimenting with autonomous frameworksexperimental autonomy
Now · in the wild
- 5AI tools run several attack stages“minimal human-in-the-loop”
Projected · forecast, not observed
- 6Simple attacks with little to no human interactionnear-term, 1–3 months
- 7Sophisticated, fully autonomous attacksmedium-term, 3–6 months
Two headlines, one report
“Have arrived”
The Decoder, Nov 16
Follows one observed case: a framework that found and exploited flaws with no human in the loop.
Evidence · one server“Coming soon”
BleepingComputer, Nov 16
Follows the timeline: stages 6 and 7 are Google’s forecast.
ForecastSolid: observed by Google. Dashed: projected by Google.
Sources: Google Threat Intelligence Group, “GTIG AI Threat Tracker”, Nov 16, 2026 (Figure 2, LLM maturity timeline); GTIG and Mandiant, Oct 30, 2026; BleepingComputer and The Decoder, Nov 16, 2026. Stage names shortened by us.
Graphic: Powered by Thorsten Meyer AI
One Case, Broader Security Stakes
The report matters because exploiting a newly patched flaw can become a race between attackers and defenders. A system that rapidly compares software versions and tests likely weaknesses could shorten the time between a patch’s release and attempts to exploit the vulnerability. For organizations, that reinforces the need to prioritize actively exploited flaws, apply security updates quickly and limit access that could help an intruder move through a network.
But the scale of the risk should not be confused with the scope of the evidence. Google described one exposed server and one framework; it did not provide a count of successful intrusions, affected organizations or attacks attributable to the system. The report’s examples of other AI tools used in real intrusions add evidence of adoption, but do not establish that those operations ran end to end without people.
Google also says current tools can trade stealth for speed: automated activity may produce bursts or patterns that are easier to spot. That is the company’s characterization, not a guarantee that every automated operation will be noisy or detected. Defenders still need to investigate unusual activity rather than rely on a single expected pattern.
Products Worth Considering

DJI Osmo 360 Motorcycle Handlebar POV Combo

Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation

Logitech MX Master 4 Ergonomic Wireless Mouse with Haptic Feedback, Graphite
As an affiliate, we earn on qualifying purchases.
Observed Activity Versus Forecasts
Google’s report is a quarterly assessment of how criminal groups, state-backed operators and influence campaigns use large language models. GTIG says its findings draw on Mandiant incident response work, threat-actor tracking and platform defenses. It places current activity at stage five of a seven-stage autonomy scale, described as “minimal human-in-the-loop.”
The report distinguishes that observed stage from what it expects next. Google projects attacks with “little to no human interaction” in the near term, defined as one to three months, and fully autonomous attacks in the medium term, defined as three to six months. Those are forecasts. In an October 30 post, GTIG described a later stage in which a person sets the target and objective while multi-agent systems carry out the attack chain; a further stage adds greater stealth and sophistication.
This distinction helps explain differing headlines about the report. Describing automated exploitation as having arrived refers to Google’s exploit_agent finding. Saying fully automated attacks are coming soon refers to Google’s projections about wider, more autonomous operations. The report supports the first as a specific observed case and presents the second as an expectation, not a confirmed trend.
“In the past quarter, GTIG identified threat actors using LLM frameworks to fully automate vulnerability discovery and exploitation without a human in the loop for the first time.”
— Google Threat Intelligence Group
What the Server Does Not Establish
Google has not publicly identified who operated exploit_agent, how many victims were affected or whether the stored victim data corresponds to successful compromises by this framework. It also does not say whether the WordPress plugin analysis shown in the report led to an exploit. The available account does not establish how long the server operated or whether it remains accessible.
The finding does not show how often attackers can reproduce this process, whether other groups have comparable systems, or whether the framework can reliably carry out a complete intrusion without human direction. Google’s stages for future autonomy are projections with stated time windows; the report does not confirm that those milestones will occur on schedule.
Defenses and Further Evidence
Google’s report advises organizations to prioritize patches for actively exploited vulnerabilities, reduce the use of long-lived passwords and keys, restrict administrative access and monitor for bursts of activity that may indicate automation. These are defensive measures; the report does not link them to a single confirmed attack pattern that all organizations can expect.
The next evidence to watch for is whether Google or other investigators identify additional cases, publish further details about exploit_agent, or document successful compromises tied to it. For now, the clearest distinction remains between one reported instance of automated exploitation and Google’s forecast that attacks requiring little or no human interaction could become more capable in the months ahead.
Key Questions
What did Google say it found?
Google says it found an exposed server running exploit_agent, a multi-agent framework that analyzed software updates, developed attacks against inferred vulnerabilities and executed them automatically.
Does the report prove fully autonomous cyberattacks are widespread?
No. It documents one observed framework and discusses other AI tools used in intrusions. Google’s prediction that fully autonomous attacks may emerge in three to six months is a forecast, not evidence that such attacks are already widespread.
Who ran the server, and how many victims were affected?
Google did not name the operator or report a number of affected websites. It said data on the server suggested interest in French media, entertainment and education sites, but that does not establish how many were successfully compromised.
What should organizations do now?
Google recommends prioritizing patches for actively exploited flaws, limiting long-lived credentials, restricting administrator access and monitoring for unusually rapid bursts of activity.
Source: ThorstenMeyerAI.com
