Can AI Carry Out Attacks Without Human Input? What Google Found And Forecast
AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Google’s threat researchers say they found an exposed server running an AI framework that automatically analyzed software patches, developed exploits and executed them. The finding documents one case of automated exploitation; Google’s predictions of broader attacks with little or no human input are forecasts, not observed events.

Google Threat Intelligence Group says it found an exposed server running an AI framework that automatically analyzed software updates, developed exploits for the vulnerabilities they appeared to fix, and launched those exploits. The case marks a documented instance of automated vulnerability discovery and exploitation, while Google’s forecast that attacks will become broadly autonomous remains a prediction.

Before you orderOffer from Amazon

Get everyday essentials 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.

In its November 16 AI Threat Tracker, produced with Mandiant, Google said the framework was called exploit_agent. It combined multiple AI agents with commercial and locally run language models. According to Google, it monitored popular web software, compared code before and after security updates, inferred which weaknesses had been fixed, then generated and executed attacks without human approval at each step.

Google said the server also contained stolen victim data and scripts for later actions, including searching within 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 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 an attack.

The report also describes attackers using open-source AI penetration-testing frameworks in real intrusions. Google said it observed CAIRN used against targets that included the US aerospace sector, and a customized PentAGI platform used against organizations in Asia. These examples show AI tools being used in attack operations, but they are distinct from the specific exploit_agent finding.

At a glance
reportWhen: Published November 16, 2026; broader au…
The developmentGoogle Threat Intelligence Group reported finding a server that automated vulnerability discovery and exploitation without a person approving each step.
Status board: what Google has seen, and what it only expects, from Google's threat report of November 16, 2026. Observed, highlighted as a first per Google: exploit_agent, one exposed server that found flaws, wrote exploits and ran them with no human in the loop. Also observed: AI attack frameworks in real intrusions, CAIRN against US aerospace targets and PentAGI in Asia; a PaperCut campaign with at least 440 installations at 395 organizations in 48 countries, per GreyNoise; stolen AI access with fraudulent bills above $100,000 in a single day; coding-assistant settings changed to download a backdoor. Forecast by Google: in 1–3 months, simple attacks with little to no human interaction; in 3–6 months, sophisticated, fully autonomous attacks. Not known: how many attacks already run fully on their own; what exploit_agent damaged and who runs it; from which date the forecast counts
What Google observed and what it only forecasts. Sources: Google Threat Intelligence Group, AI Threat Tracker, November 16, 2026; GTIG and Mandiant, October 30, 2026; GreyNoise, PaperCut campaign report. Graphic: Thorsten Meyer AI.

A Faster Route From Patch to Exploit

Software updates can reveal clues about the vulnerabilities they repair. A system that rapidly compares pre-update and post-update code, then tests and launches an exploit, could give attackers a way to act sooner after a patch becomes available. That raises the pressure on organizations to apply fixes quickly, especially for flaws already known to be under active exploitation.

Google also said current tools can trade stealth for velocity: automated activity may be fast but noisy, making unusual bursts easier for defenders to spot. The report’s evidence supports concern about faster attack workflows, but it does not show that fully autonomous attacks have become routine or that the observed framework succeeded against a stated number of victims.

Products Worth Considering

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite

Sony WH-1000XM5 Wireless Noise-Canceling Over-Ear Headphones, Black, International Version

Sony WH-1000XM5 Wireless Noise-Canceling Over-Ear Headphones, Black, International Version

As an affiliate, we earn on qualifying purchases.

Google’s Seven-Stage Timeline

Google places attacker use of AI on a seven-stage scale. The first four stages cover earlier uses, from prompting chatbots to experimenting with autonomous frameworks. Google labels the current stage, stage five, as tools running several parts of an attack with minimal human involvement.

Stages six and seven are projections. Google expects simple attacks with little to no human interaction in the near term, defined as one to three months, and sophisticated fully autonomous attacks in the medium term, three to six months. The tracker draws on Mandiant incident-response work, threat-actor tracking and Google platform defenses. Its central claim about fully automated vulnerability discovery and exploitation is based on the exposed server it identified.

News coverage used different timing in its headlines. The Decoder focused on the observed case and described fully automated attacks as having arrived; BleepingComputer emphasized Google’s forecast for more autonomous attacks and described them as coming soon. The distinction is that Google reports one observed automated exploitation framework, while its timeline for wider autonomy is forward-looking.

“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 did not say who operated exploit_agent, how many websites were successfully compromised, or whether the WordPress plugin checks shown in the report led to exploitation. The server’s data suggested interest in French media, entertainment and education websites, but Google did not provide a confirmed victim count.

The report documents an automated framework in one case. It does not establish how common such systems are, whether they can reliably carry out complex attacks without people, or when Google’s projected stages six and seven will arrive. The one-to-three-month and three-to-six-month windows are forecasts, not confirmed milestones.

Defenders Face a Shorter Response Window

Google’s report points organizations toward practical defenses: prioritize patches for actively exploited vulnerabilities, reduce long-lived passwords and keys, restrict administrator access, and monitor for bursts of activity that may indicate automated operations. These steps address the faster attack cycles described in the report.

The next developments to watch are further evidence about how often attackers use frameworks like exploit_agent, whether investigators identify its operator or confirmed victims, and whether attacks with little human interaction appear within Google’s forecast windows. The report does not name a specific next release or date for an updated timeline.

Key Questions

What did Google find?

Google said it found an exposed server running exploit_agent, an AI framework that automated software vulnerability research and exploit execution without human approval at each step.

Does the report prove fully autonomous attacks are widespread?

No. Google documented an automated framework in one case. Its forecast that simple attacks with little human interaction will emerge in the near term, followed by more sophisticated autonomous attacks, is a projection.

Who was targeted by the exploit_agent operator?

Google said data on the server suggested interest in French websites in media, entertainment and education. It did not identify the operator or give a confirmed number of compromised sites.

What should organizations do in response?

Google recommends prioritizing patches for actively exploited flaws, limiting long-lived credentials, restricting administrator access and watching for unusually rapid bursts of activity.

Source: Thorsten Meyer AI

You May Also Like

Regulatory Sandboxes: Safe Havens or Security Nightmares for AI Testing?

In exploring regulatory sandboxes for AI testing, understanding how to balance innovation with security concerns is crucial—are they truly safe havens or potential nightmares?

Notes, Handoffs, And Permission: Inside AI Agent Swarms

A researcher report and an OpenAI evaluation show how agents pass information, work around blocks and mistake peer messages for authorization.

How AI Security Maturity Differs From General Cyber Maturity

Guiding you through the nuances, discover how AI security maturity uniquely evolves to address sophisticated threats beyond traditional cyber defenses.

Why 4K NVR Security Systems Appeal to Small Businesses

AIThis post was created with the assistance of artificial intelligence (AI).4K NVR…