Google: Fully Automated AI Cyberattacks Have Arrived
Google’s Threat Intelligence Group said on Nov 16 that attackers have for the first time fully automated vulnerability discovery and exploitation with LLM frameworks, with no human in the loop; near-human-free attacks are expected within three months.

What happened
Google’s Threat Intelligence Group (GTIG) and Mandiant published their quarterly AI Threat Tracker on Nov 16, titled “Agentic Threat Activity Crosses a Significant Threshold.” The report says that in the past quarter, attackers used LLM frameworks to fully automate vulnerability discovery and exploitation without a human in the loop for the first time, and that financially motivated and state-sponsored groups are both adopting agentic workflows at scale.
The key evidence comes from an attacker’s exposed staging server, where GTIG found a multi-agent vulnerability research and weaponization framework dubbed exploit_agent, integrating multiple frontier and local LLMs. The framework continuously compares the code of popular web applications before and after patches, identifies the vulnerability each patch fixed, writes an exploit and runs it automatically; its targets were French media, entertainment and education websites. The server also held stolen victim data, reconnaissance information, webshells and brute-forcing scripts; configuration files were in Russian, and Google did not attribute the operator.
The report also names open-source pentesting frameworks now systematically abused: ARTEX, CAIRN, CyberStrikeAI, HexStrike AI and PentAGI. CAIRN was used in an attack against the US aerospace sector; suspected North Korean hackers used Claude Code to deliver malware; the UNC6836 cluster attacked crypto firms in North America and Europe with a modified Claude Code, changing its settings after the initial compromise so it downloaded the SOMBERMEME backdoor at launch. In the PHP2Shell episode, hundreds of purported proof-of-concept exploits, mostly AI-generated, appeared shortly after PHP disclosed two flaws.
Timeline and outlook
Google estimates that attacks requiring little to no human interaction and relying mostly on multiple AI agents will appear within three months, and sophisticated, fully autonomous agentic attacks within six months. Current attacks trade stealth for speed, leaving forensic traces that make them easier to detect, but deploying such frameworks has dramatically shortened the time defenders have to patch.
Context: the FBI warned last week that multiple threat actors have weaponized the ARTEX framework since August, attacking the financial, government, technology, defense and healthcare sectors (covered in this site’s Nov 15 flash). Google rates current activity at Level 5, where AI handles most stages of an attack while a human may only set it up.
Key facts
- First: In the past quarter, LLM frameworks fully automated vulnerability discovery and exploitation without a human in the loop for the first time
- exploit_agent: A multi-agent framework integrating multiple frontier and local LLMs; compares code before and after patches, writes exploits and executes them automatically
- Targets: French media, entertainment and education websites; the server held stolen data, webshells and brute-forcing scripts; configuration files were in Russian and the operator is unattributed
- Tooling: ARTEX, CAIRN, CyberStrikeAI, HexStrike AI and PentAGI are now systematically abused
- Outlook: Near-human-free attacks within three months; sophisticated fully autonomous attacks within six months
- Status: Level 5: AI handles most stages of an attack while a human may only set it up; speed is traded for stealth, leaving attacks easier to detect
- Related: CAIRN targeted the US aerospace sector; suspected North Korean hackers used Claude Code to deliver malware; UNC6836 attacked crypto firms with a modified Claude Code
Why it matters
Vulnerability discovery and exploitation have gone fully human-free for the first time, sharply shrinking the window between a patch release and its exploitation; Google rates current activity at Level 5 and expects stealthier fully autonomous attacks within six months.
CompaniesGoogle
Grab todayOpenAI Codex for Students: $100 in Free CreditsHIMIAO Picks →Follow HIMIAO Horizon
No sign-up · Unsubscribe any timeNotifications
A ping on your phone or computer when the brief is out. No email needed.
Comments
Related
- Nov 7DeepMind Agents Find Two Weaknesses in Legacy 1024-Bit RSA
- Oct 19MONST3R-AI Crimeware Runs Autonomous Pentest Scans on Stolen API Keys, Victims Foot the Bill
- Oct 20Google Open-Sources Harness²: Letting AI Agents Rewrite Their Own Scaffolding, Hits GitHub Trending
- Nov 15Google Project Zero Rewrites Disclosure Policy: 90 Days Now, 30 Days by 2028
- Nov 13DeepMind Formally Verifies Signal's Crypto Protocol With 6.5M Lines of Proof
Today
Nov 17 Tuesday- Brief
- GitHubGitHub Weekly Vol.6: 7 Projects Worth a Look This Week
Tested or read — weeklyAll →
- Quiz
- Call
Latest news
All →- Nov 17Moonshot Pivots Kimi to Science: Biology, Chemistry and Pharma
- Nov 17OpenAI Launches CSAM Image Classifier With 24-Hour ChatGPT Suspensions
- Nov 17US Commerce Department Names 33 AI Export Packages, AMD Leads 11
- Nov 17US Commerce Department Closes "Cloud Loophole" for Advanced AI Chips
- Nov 17OpenAI Launches Science Blog as GPT-6 Astra Leads Proof of Candidate Lattice Standard Model Definition
- Nov 17Apple Smart Glasses Said to Debut at WWDC 2027 With No Display, Leaning on Siri and iPhone
™