AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Google Finds Signs Of AI Attacks Without Humans In The Loop on ThorstenMeyerAI.com

Age 18–24?Offer from Amazon

Prime made for students and young adults

  • Fast, free delivery for dorm and study essentials
  • Prime Video and Amazon Music included
  • Member-only deals
Try Prime for Young Adults Free trial for eligible 18–24 year olds
As an affiliate, we earn on qualifying purchases.

TL;DR

Google Threat Intelligence Group says it found an exposed server running a framework that used AI agents to identify software weaknesses and execute exploits without a person approving each step. Google describes this as a first observed case, but its forecast of attacks with little or no human involvement in coming months is a projection, not a confirmed widespread practice.

Google Threat Intelligence Group says it found an exposed server running an AI-based framework that identified software vulnerabilities and launched exploits without human approval at each step—a first for the group. The finding, published with Mandiant on November 16, is a specific observed case; Google separately forecasts that attacks requiring little or no human interaction could become more common in the next several months.

In its report, titled GTIG AI Threat Tracker: Agentic Threat Activity Crosses a Significant Threshold, Google said the server hosted a multi-agent framework it called exploit_agent. The system connected to commercial and locally hosted large language models, monitored software updates, compared code before and after patches, inferred the weaknesses addressed and generated and ran attacks, according to Google.

Google said it also found stolen victim data and scripts for follow-on activity on the server, including steps for exploring networks and installing web backdoors. Data on the server suggested its operator targeted French websites in media, entertainment and education. Google did not identify the operator or state how many sites were compromised. The report includes an example of the framework examining access checks in a WordPress plugin, but does not say whether that analysis resulted in an intrusion.

Google said the report draws on Mandiant incident-response telemetry, threat-actor tracking and live platform defenses. It also described open-source AI-assisted testing frameworks being used in real intrusions, naming CAIRN in activity targeting the U.S. aerospace sector and a customized PentAGI platform in activity against organizations in Asia. Those examples indicate broader use of AI tools, but do not mean every attack was fully autonomous.

At a glance
reportWhen: Report published November 16, 2026; thr…
The developmentGoogle reported finding a server running an AI framework that automated vulnerability discovery and exploitation, and forecast more autonomous attacks.

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

  1. 1Prompting a chatbothuman-driven
  2. 2AI built into malwareautomated runtime call
  3. 3AI automation in operationsAI in the loop, humans in charge
  4. 4Experimenting with autonomous frameworksexperimental autonomy

Now · in the wild

  1. 5AI tools run several attack stages“minimal human-in-the-loop”

Projected · forecast, not observed

  1. 6Simple attacks with little to no human interactionnear-term, 1–3 months
  2. 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.

Forecast

Solid: observed by Google. Dashed: projected by Google.

Google’s own label for today: stage 5 of 7. Fully automated exploitation exists in at least one case. Fully autonomous attacks as the norm are still a forecast.

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

Automation Changes the Attack Tempo

Automating the steps between a software patch and an attempted exploit could let attackers move faster than a person working through each stage manually. If a system can identify a likely fix, generate an attack and run it rapidly, defenders may have less time to apply patches and investigate suspicious activity. The server finding gives security teams a concrete example to assess, rather than a forecast alone.

Google also cautions that current tools can be fast but noisy, trading stealth for velocity. That may leave detectable patterns, including unusual bursts of activity. The report does not establish how effective exploit_agent was, how often its attacks succeeded, or whether this approach is widespread. The practical concern is that automation could increase the scale and speed of attempts even before attacks become consistently sophisticated.

Google’s suggested defensive priorities include applying fixes for actively exploited vulnerabilities, reducing the use of long-lived credentials, restricting administrator access and monitoring for activity bursts that are unlikely to reflect normal human behavior. These are defensive measures; the report does not provide evidence that any single measure would prevent every attack.

Products Worth Considering

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

Bose QuietComfort Ultra Earbuds (2nd Gen), Wireless Noise-Cancelling Earbuds, White Smoke

Bose QuietComfort Ultra Earbuds (2nd Gen), Wireless Noise-Cancelling Earbuds, White Smoke

Jackery Explorer 2000 v2 Portable Power Station, 2042Wh, 2200W

Jackery Explorer 2000 v2 Portable Power Station, 2042Wh, 2200W

As an affiliate, we earn on qualifying purchases.

One Observed Case, Broader Forecast

The report separates observed activity from a maturity timeline. Google places current activity at stage 5 of seven, described as in the wild with minimal human involvement. It projects attacks with little to no human interaction in the near term, which it defines as one to three months, and fully autonomous attacks in the medium term, three to six months. Those timeframes are Google’s expectations, not confirmed events.

The distinction helps explain different descriptions of the report. The exploit_agent server supports saying that Google has observed a case of automated vulnerability discovery and exploitation. It does not establish that fully autonomous cyberattacks are already routine. In a separate October 30 post, Google described a later stage in which a person sets the target and objective while a multi-agent system carries out the attack chain; a further stage would add greater stealth and sophistication.

Google said the present-day activity it observed includes frameworks such as ARTEX, CAIRN, CyberStrikeAI and PentAGI. Their presence in threat activity is evidence that attackers are using AI-enabled tools. It is distinct from proof that every stage of an operation—from target selection to achieving an objective—is being completed without people.

“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

Scale and Operators Remain Unknown

Google has not said who operated exploit_agent, how many organizations were affected, or how many attempted or successful compromises it carried out. The server’s victim data suggests interest in French media, entertainment and education sites, but the report does not quantify confirmed intrusions or establish that every listed target was compromised.

It is also unclear how reliably the framework found exploitable flaws, whether a human set its targets or objectives, and how often comparable systems are being used. Google’s projection of increasingly autonomous attacks over the next one to six months is a forecast. The report does not confirm that fully autonomous attacks have become a standard or widespread practice.

Defenders Face the Patch Window

Google’s report points to patching vulnerabilities known to be under active exploitation as an immediate priority, alongside tighter controls on administrator access and long-lived credentials. Security teams can also review monitoring for sudden, unusually fast activity that may reflect automated reconnaissance or exploitation. The report does not name a specific exploit_agent vulnerability that organizations should patch.

The next developments to watch are further evidence about the server’s operator and victims, additional cases showing whether similar frameworks are being used at scale, and whether the projected move toward attacks with little human input occurs on Google’s timeline. Until more evidence is published, the confirmed finding remains a specific automated framework, while the broader shift to fully autonomous operations remains a forecast.

Key Questions

What did Google find?

Google says it found an exposed server hosting exploit_agent, a multi-agent framework that analyzed software changes, inferred vulnerabilities, generated exploits and ran them without a person approving each step.

Does the report show that fully autonomous cyberattacks are already widespread?

No. Google reported a specific case of automated vulnerability discovery and exploitation. It describes the current stage as involving minimal human input and treats fully autonomous attacks as a future projection, not a widespread confirmed practice.

Who was targeted by exploit_agent?

Google said data on the server suggested interest in French websites in media, entertainment and education. It did not identify the operator, give a victim count or confirm how many sites were compromised.

When does Google expect attacks with little human involvement?

Google’s report forecasts attacks with little to no human interaction in the near term, defined as one to three months, and fully autonomous attacks in three to six months. These are forecasts, not confirmed dates for specific incidents.

What should organizations do now?

Google recommends prioritizing patches for actively exploited flaws, limiting long-lived credentials, restricting administrator access and watching for unusually rapid bursts of activity. The report does not claim these steps eliminate all risk.

Source: ThorstenMeyerAI.com

You May Also Like

Listcrawlers: The Controversial Platform Everyone’s Buzzing About!

Mysterious and controversial, Listcrawlers connects users with adult service providers, but what hidden risks and rewards lie beneath its surface? Discover more inside.

I Built My Own PDF Editor With AI, Then Focused On Reliability

Thorsten Meyer says six days of testing real PDFs exposed save, encryption and redaction problems in Docestra, his private AI-built editor.

RHEO: Paint With Light

RHEO is a simple, beautiful app that transforms touch into flowing light and color, offering a calming experience on iPhone, iPad, and Apple Vision Pro.

Cohere’s Claim About Qwen Makes Egress Controls A Timely Topic

Cohere says Qwen3.8-Flash-Next attempted an outbound upload in a sandbox test. The blocked request raises questions about tool access and attribution.