Google Identifies First Fully Autonomous AI Cyberattack Targeting French Sectors

3 min read
Source: CyberScoop
TL;DR

Google’s Threat Intelligence Group reported the first fully autonomous cyberattack, where AI agents identified and exploited software vulnerabilities without human intervention. The operation, attributed to Russian-speaking hackers, targeted French media and education sectors. This marks a shift to 'machine speed' attacks, compressing the window between patch disclosure and exploitation. Google is responding with AI-driven defense tools to match this new threat landscape.

Key points

  • Google’s Threat Intelligence Group and Mandiant identified the first cyberattack run end-to-end by AI, with no human involvement in the intrusion process.
  • The attack, dubbed 'exploit_agent,' used AI to compare code before and after vendor patches, infer flaws, draft exploits, and execute them automatically.
  • Attribution rests on Russian-language comments in the code, including instructions for 'auto-exploitation.' Victim data suggests targeting of French websites in media, entertainment, and education.
  • Google observed open-source frameworks like CAIRN and PentAGI being used against an aerospace organization in the US and targets in Asia.
  • The number of exploited zero-days rose to 22 in August 2026, up from 11 earlier in the year and an average of eight in 2025.
  • Google is launching AI Threat Defense, combining Gemini, Wiz, CodeMender, and Mandiant to automate vulnerability detection and patching.

Background

This development follows Google’s September report noting 'nascent efforts' in AI-augmented vulnerability research. It aligns with broader concerns over autonomous AI systems, including Anthropic’s release of Mythos and an OpenAI agent incident. Recent archive coverage highlights increased Russian-linked cyber activities, including data leaks and watering hole campaigns targeting diplomats.

How outlets are covering it

CyberScoop and UPI emphasize the novelty of fully autonomous attacks and the Russian-language attribution. The Hacker News highlights the broader trend of using open-source frameworks like ARTEX and CyberStrikeAI, noting espionage and financially motivated actors. Google’s blog focuses on the defensive response, stressing the need for 'agentic defense' to match machine-speed threats. All sources agree on the shift to automated exploitation but differ in emphasis: some focus on attribution, others on the technical mechanics or defensive solutions.

Why it matters

The emergence of fully autonomous AI cyberattacks represents a paradigm shift in cybersecurity, compressing the window between vulnerability disclosure and exploitation. This increases the risk of widespread, rapid intrusions, particularly for organizations in targeted sectors. It necessitates a move from manual defense to automated, AI-driven security measures to keep pace with machine-speed threats.

What to watch

Google expects threat actors to continue using these frameworks in global campaigns due to low barriers to entry. The company is rolling out AI Threat Defense to automate patching and vulnerability detection. Defenders must adopt agentic security tools to match the speed and scale of AI-driven attacks, which are projected to become widespread within six months.

Share this article

Want the full story? Read the original reporting

Read on CyberScoop