Google projects fully autonomous AI cyberattacks within six months

RUKUS· · 3d ← 5d
Go Rogue

Google's Threat Intelligence Group warns that autonomous AI cyberattacks have arrived, with AI agents scanning software updates and building attacks without human involvement.

Its AI Threat Tracker places current attacker activity at Stage 5, with AI models running several stages of an attack with at least one human operator involved.

Google projects fully autonomous agentic attacks within three to six months, after simpler multi-agent attacks with little human interaction within one to three months.

Google says attackers are not using new break-in methods but exploiting known, unfixed flaws with far greater speed and scale than human defenders can match.

Google is countering with AI Threat Defense, which combines Gemini, Wiz, the CodeMender patching agent, and Mandiant expertise.

Google says it confirmed the first cyberattack carried out end to end by AI, from finding a vulnerability through preparation to execution, without human intervention.

  1. Google announced on November 15, 2026 that it had confirmed for the first time a cyberattack completed end to end by artificial intelligence, from identifying a vulnerability through preparation to execution.

    An analysis of the attacker's server found that autonomous AI agents fully automated the attack without human operation or intervention.

  2. On November 17, 2026, Google's Threat Intelligence Group placed the threat from autonomous AI models at Stage 5, meaning AI models had carried out multiple attack lifecycle stages with at least one human operator involved.

    The group's AI Threat Tracker report says bad actors are using autonomous LLM frameworks to continuously scan software updates, work out what flaw was patched, and build an attack for it without human involvement.

    Google announced Google AI Threat Defense, which combines Gemini and other frontier models, Wiz's prioritization, CodeMender's automated patching, and Mandiant's expertise to find and fix flaws before attackers exploit them.

    The Threat Intelligence Group flagged cyber espionage and impacts on supply chain systems among trends in incidents involving AI models in the months before the warning.

  3. Google's report says AI agent frameworks run around the clock without stopping like a human would, scanning and attacking targets at speeds human IT and security teams cannot keep up with, leaving no time to manually review and patch systems.

    Google Threat Intelligence Group chief analyst John Hultquist said, "For the most part, attackers aren't using new groundbreaking methods to break into systems – they are still going after known, unfixed issues, just with much greater speed and scale".

    Hultquist said, "Instead of relying on manual effort to keep up, we will have to embrace agentic defense. Or we risk falling behind".

  4. The Google Threat Intelligence Group and Mandiant's AI threat tracker projects fully autonomous agentic attacks within three to six months, following multi-agent attacks with little or no human interaction within one to three months.

    The tracker says threat actors used open-source AI penetration-testing frameworks named ARTEX, CAIRN, CyberStrikeAI, and PentAGI, with CAIRN targeting the U.S. aerospace sector.

    The tracker describes an exploit_agent system found on an exposed staging server that compared code before and after patches, wrote exploits, and ran them without a human in the loop, targeting flaws that already had patches.

    Google says the AI attack tools trade stealth for speed and leave tracks behind, which helps detection businesses such as Palo Alto Networks' Cortex and Unit 42.

    The Google Threat Intelligence Group and Mandiant forecast fully autonomous agentic attacks within three to six months, after multi-agent attacks with little to no human interaction within one to three months.

Sign in with subscription

Free during the beta.

Clear all notifications?

Your bells stay on. Only this list empties.