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

🔍 Read the full analysis: AI Attacks Without Human Input? Google Distinguishes Findings From Forecasts on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Google Threat Intelligence Group says it found an exposed server running an AI-agent framework that could identify vulnerabilities from software patches, write exploits and execute them without human approval at each step. Google describes this as an observed case, but says attacks with little or no human interaction more broadly are projected for the coming months, not established as routine.

Google Threat Intelligence Group (GTIG) says it found an exposed server running an AI-agent framework that could analyze software updates, infer the vulnerabilities they fixed, create exploits and run them without a person approving each step. The November 16 report calls this the first instance the group has identified of attackers fully automating vulnerability discovery and exploitation, while treating wider, fully autonomous attacks as a forecast rather than a current norm.

In its report, GTIG AI Threat Tracker: Agentic Threat Activity Crosses a Significant Threshold, produced with Mandiant, Google describes a framework it calls exploit_agent. The system was hosted on an exposed staging server and connected to multiple commercial and locally run large language models. According to Google, it monitored widely used software, compared code before and after security patches, inferred what weaknesses the patches addressed, drafted exploits and automatically executed them.

Google also says the server contained stolen victim data and scripts for follow-on activity, including network reconnaissance and installing web backdoors. Data on the server suggested an interest in French websites in media, entertainment and education. Google did not identify the server’s operator, report how many sites were compromised or say whether a pictured analysis of a WordPress plugin’s access checks resulted in an attack.

The report describes other AI-assisted tools in real intrusions, including frameworks named CAIRN and PentAGI. Google says it observed CAIRN used against targets including the U.S. aerospace sector and a customized PentAGI platform used against organizations in Asia. Those examples support Google’s account of growing AI use in cyber operations, but they are distinct from its specific finding that exploit_agent automated vulnerability discovery and exploitation.

At a glance
reportWhen: Published November 16, 2026; Google’s f…
The developmentGoogle’s November 16 threat report describes an observed framework that automated vulnerability discovery and exploitation, while forecasting more autonomous attacks in the months ahead.

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

What Automation Changes for Defenders

Automating the path from a software patch to a working exploit could compress the time defenders have to update vulnerable systems. A tool that can inspect code changes and act on its own may let an operator pursue more targets or move faster than a workflow requiring a person to approve each stage. The report does not establish how widely this particular framework was used, but the observed capability gives security teams a reason to watch for rapid exploitation after patches are released.

Google also cautions that speed comes with a trade-off. It says current tools “trade stealth for velocity”: their rapid activity can be noisy and easier for defenders to detect. That is a description of the tools Google has observed, not a guarantee that future systems will remain easy to spot. The practical concern is both sides of that trajectory: defenders may face faster attacks now, while increasingly capable and less conspicuous systems remain a forecast.

For organizations, Google’s findings reinforce the value of prioritizing actively exploited vulnerabilities, restricting administrator access and reducing the lifespan of passwords and keys. Security teams can also monitor for unusually dense bursts of activity that appear automated. These are defensive steps, not evidence that every organization is currently facing fully autonomous attacks.

Products Worth Considering

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

As an affiliate, we earn on qualifying purchases.

Observed Activity Versus Forecasts

Google’s tracker is a quarterly assessment of how criminal groups, state-backed actors and influence operations use large language models. GTIG says it draws on Mandiant incident-response telemetry, global threat-actor tracking and live platform defenses. The report says the threat landscape has reached a significant threshold, but its evidence and projections describe different levels of automation.

Google places current activity at stage five of a seven-stage timeline, labeled “minimal human-in-the-loop.” It projects attacks with little or 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 time windows are Google’s expectations, not confirmed dates for a general shift in attacker behavior.

The distinction helps explain differing headlines about the report. The claim that automated attacks have “arrived” refers to Google’s observed exploit_agent case. The claim that they are “coming soon” reflects Google’s projection of broader, more autonomous operations. The report supports the first as a finding in at least one case; it does not say that fully autonomous attacks are already routine.

““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

Scope of the Exploit Agent Case

Google has not said who operated the server, how many websites were affected or whether the data it found can be tied to successful intrusions. Its report describes victim data and scripts, but the supplied account does not establish the full scale or outcome of the activity. It also remains unclear whether the framework’s examination of a WordPress plugin led to exploitation.

The report’s broader projections carry uncertainty too. Google places some future stages months away, but it does not establish that the predicted capabilities will emerge on that schedule or become common. One observed framework is evidence of a capability, not proof that all attackers—or even a large number of them—are using it.

What Security Teams Should Monitor

Google’s report sets out no public timetable for a follow-up investigation into exploit_agent. The next developments to watch are whether researchers identify the operator or additional victims, and whether further incident reports show similar systems being used to discover and exploit vulnerabilities with little human involvement.

For now, organizations can use the report as a prompt to review patch prioritization, administrator permissions and credential lifetimes. Monitoring for unusually rapid bursts of activity may help identify automated behavior. Google’s predicted one-to-three-month and three-to-six-month stages should be treated as its outlook; whether those milestones occur remains to be seen.

Key Questions

What did Google say it observed?

Google says it found an exposed server running exploit_agent, a framework that could analyze patches, infer vulnerabilities, draft exploits and execute them automatically.

Does the report say fully autonomous attacks are already common?

No. Google describes one observed case of automated vulnerability discovery and exploitation. Its report labels the current stage “minimal human-in-the-loop” and presents broader fully autonomous attacks as a projection.

Who was targeted by exploit_agent?

Google says data on the server suggested interest in French websites in media, entertainment and education. It has not identified the operator or specified how many sites were affected.

What does Google recommend organizations watch for?

The report’s practical advice includes prioritizing actively exploited flaws, limiting administrative access, shortening the life of passwords and keys, and watching for bursts of activity that may indicate automation.

Source: ThorstenMeyerAI.com

You May Also Like

Bringing AI To All Languages: A Game Changer For Worldwide Connectivity

Google announces support for over 300 languages, real-time translation, and new models to include underrepresented languages, advancing global connectivity.

What’s The ROI Of Software Engineers’ Spending On AI?

A survey of 27 workplaces finds sharply different AI costs and higher PR counts in some data, but few companies can quantify the return.

A Lower Bill, A Small Gain: Assessing Google’s Gemini 4 Neon

Google’s Gemini 4 Neon scored one point above Carbon on an independent index and completed tasks faster, but leaders remain ahead.

QuantrolOx Raises €12 Million In Series A2

The Oxford spinout will use its €12 million Series A2, led by IW Capital with EIC Fund participation, to expand products, training and U.S. operations.