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

🔍 Read the full analysis: Google’s AI Attack Report: Findings Today And Forecasts For Tomorrow on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

Google’s November 16 AI Threat Tracker reports that researchers found an exposed server running a framework that analyzed software patches, developed exploits and executed them without a person approving each step. Google describes this as a first observed instance, but says attacks with little or no human involvement are a forecast for the coming months, not a description of the current norm.

Google’s Threat Intelligence Group said it found an exposed server running a system that could analyze software updates, infer the vulnerabilities they fixed, write exploits and run them without human approval at each step. The November 16 report describes this as the first time Google has observed attackers fully automating vulnerability discovery and exploitation, while warning that attacks with little or no human involvement may emerge in the coming months.

The report, published with Mandiant under the title GTIG AI Threat Tracker: Agentic Threat Activity Crosses a Significant Threshold, says the server hosted a multi-agent framework Google called exploit_agent. It connected to commercial and locally run large language models and was designed to compare software code before and after security patches, identify likely fixes, draft attacks and execute them. Google said its findings draw on incident-response work, threat tracking and live platform defenses.

Google found stolen victim data and scripts for later activity on the server, including network reconnaissance and planting web backdoors. The victim data suggested the operator was interested in French websites in media, entertainment and education. Google did not identify the operator or say how many sites were compromised. A report figure shows the framework examining access controls in a WordPress plugin, but Google did not state that this examination resulted in an attack.

The report also describes open-source AI-assisted security frameworks being used in real intrusions. Google said it observed CAIRN used against targets including the U.S. aerospace sector, and a customized PentAGI platform used against organizations in Asia. These examples are separate from exploit_agent; the report does not say they carried out the same patch-analysis and automated-exploitation process.

At a glance
reportWhen: Report published November 16, 2026; for…
The developmentGoogle’s threat researchers reported finding an AI-based framework that automated vulnerability discovery and exploitation, and forecast more autonomous attacks in the next several months.

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

A Faster Path From Patch to Exploit

Automating the steps between a software update and an attempted exploit could give attackers a way to act quickly after a flaw is fixed. A patch can reveal clues about the underlying weakness; a system that analyzes code changes and produces an attack may shorten the time defenders have to update affected software. Google’s report makes the development relevant to organizations that rely on widely used web software and need to prioritize security updates.

But the distinction between an observed capability and a broad change in attacker behavior matters. Google described one exposed server running exploit_agent, not evidence that fully automated exploitation is widespread. The report also says current tools trade stealth for velocity: their activity can be fast but noisy, potentially giving defenders signals to detect. The practical concern is a changing balance between the speed of attacks and the time organizations take to patch and respond.

Breville BES870BSXL Barista Express Espresso Machine, Black Sesame

Breville BES870BSXL Barista Express Espresso Machine, Black Sesame

De'Longhi Stilosa Compact Manual Espresso Machine with Milk Frother

De'Longhi Stilosa Compact Manual Espresso Machine with Milk Frother

As an affiliate, we earn on qualifying purchases.

Observed Automation, Projected Autonomy

Google’s report places current activity at stage five of a seven-stage maturity timeline, labeled “minimal human-in-the-loop.” It identifies the next stage, with little to no human interaction, as a near-term projection of one to three months, and fully autonomous attacks as a medium-term projection of three to six months. Those intervals are Google’s forecasts, not confirmed deadlines or proof that such attacks will become routine.

That distinction helps explain different headlines about the report. Coverage describing automated attacks as having arrived refers to the exploit_agent case Google says it observed. Coverage saying they are coming soon refers to the later stages in Google’s timeline, which remain projected. In an October 30 post, GTIG described a later-stage model in which people set the target and objective while multi-agent systems carry out the attack chain. The November report says the present stage still involves some human participation.

““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 and Operators Remain Unknown

Google has not disclosed who operated exploit_agent, how many organizations were affected, or whether the server’s examination of the WordPress plugin led to a successful intrusion. The report’s description of French-sector interests is based on victim data found on the server; it does not establish the full set of targets or the operator’s identity.

It is also unclear how often attackers can use this approach successfully or whether it will spread beyond the examples Google describes. The one-to-three-month and three-to-six-month windows are forecasts. Google’s findings do not establish that fully autonomous attacks are now common, or that the projected stages will arrive on that schedule.

Defenders Face Shorter Response Windows

Google’s report points to immediate defensive priorities: patch actively exploited vulnerabilities first, reduce the use of long-lived passwords and keys, restrict administrative access and monitor for unusually rapid bursts of activity. These steps address risks described in the report, although Google did not say they would prevent every AI-assisted intrusion.

The next developments to watch are evidence of similar systems in other investigations and whether attackers move from limited human supervision toward the later stages in Google’s timeline. The report gives no specific follow-up date or named milestone. Its forecasts should be treated as expectations to monitor, while the exploit_agent finding remains the concrete example Google says it has observed.

Key Questions

What did Google say it found?

Google said an exposed server hosted exploit_agent, a framework that could analyze software changes, infer vulnerabilities, draft exploits and execute them without human approval at each step.

Does the report show that fully autonomous cyberattacks are common?

No. Google described an observed automated-exploitation case, but its report labels the current stage “minimal human-in-the-loop”. Fully autonomous attacks are a projection, not a claim that they are already the norm.

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 or provide a total number of affected websites.

When does Google expect more autonomous attacks?

Google’s timeline projects attacks with little to no human interaction within one to three months and fully autonomous attacks within three to six months. These are forecasts, not confirmed dates.

What steps does the report recommend defenders prioritize?

The source recommends prioritizing patches for actively exploited flaws, reducing long-lived credentials, restricting administrator access and watching for unusually rapid activity.

Source: ThorstenMeyerAI.com

You May Also Like

Customer service + BPO. The operational-scale displacement.

Empirical evidence shows customer service and BPO sectors are experiencing widespread AI-driven workforce displacement, with a shift to hybrid models.

The SSD Squeeze: Why Storage Joined The Party

Enterprise and consumer SSD prices soar as NAND supply tightens due to AI workloads and wafer competition, impacting markets and buyers worldwide.

Is DoorDash down? Thousands report errors amid widespread outage; ‘something went wrong’ | Hindustan Times

Thousands of users report errors as DoorDash experiences a major outage, disrupting food delivery services nationwide. The cause remains unclear.

A New Sign-In Step On Reddit: What’s Driving The Change?

Reddit plans login tests outside the U.S. in Q4 2026, citing falling search referrals, data scraping and ad targeting. The rollout details remain unsettled.