Google Reveals How Hackers Are Using AI Agents to Automate Real-World Cyberattacks
What happened?
Hackers are no longer just asking AI chatbots for help. They now run AI agents that can scan, break in, and move through networks with very little human input. In its latest AI Threat Tracker, published on November 16, 2026, the Google Threat Intelligence Group (GTIG) and Mandiant said AI misuse has reached a “critical inflection point” across cybercrime, spying, and influence campaigns. The findings come from Mandiant incident response cases, global tracking of threat groups, and Google’s own platform defenses.
Key details
- Google Threat Intelligence Group and Mandiant report AI misuse has reached a critical inflection point across cybercrime, espionage, and influence operations.
- Attackers are using open-source multi-agent hacking frameworks such as ARTEX, CAIRN, CyberStrikeAI, and PentAGI to automate intrusions.
- CAIRN was used in real intrusions against US aerospace organizations, while a custom PentAGI version scanned targets across Asia in late August to early October 2026.
- An AI framework called exploit_agent was found on an exposed staging server that autonomously finds bugs, writes exploits, and runs them without human involvement.
- After two PaperCut zero-day flaws were exploited in August 2026, an AI-assisted campaign hit hundreds of organizations and reached domain admin within minutes.
- Russian-linked threat actors steal AI keys and cloud access, reselling them through LLM Mixers with fake charges exceeding $100,000 in a single day.
- North Korean cluster MIDNIGHT NEPTUNE used AI-generated recruiter profiles and ClickFix lures to breach crypto firms, then modified Claude settings to deploy SOMBERMEME backdoor.
What is affected?
- US aerospace sector organizations targeted by CAIRN framework.
- Organizations across Asia scanned by customized PentAGI between late August and early October 2026.
- French media, entertainment, and education websites targeted by exploit_agent framework.
- IT and software vendors targeted via Okta, Entra, and Cognito identity systems by ARTEX-using actor.
How do I fix or mitigate it?
- For AI services, set budget alerts, scan code for leaked keys, and treat unexpected AI service activation as a serious warning.
What should defenders look for?
- Noisy traffic, known tool names, and clear code traces left by current AI agent frameworks.
- Sudden spikes in API and identity activity.
- Unexpected AI service activation.
- Fake charges on AI services exceeding $100,000 in a single day.
- Modified .claude/settings.json files to download SOMBERMEME backdoor at startup.
- GitHub tokens named after Microsoft Copilot used to hide tracks.
Summarised by SecOraLens from Cyber Security News. Always read the original source before acting. How SecOraLens verifies advisories.
What to do next
Brief your board. Brief your team.
Executive risk briefs for governance, technical kill-chains for the SOC — delivered as advisories publish. Critical alerts within 1 hour. No spam, unsubscribe anytime.
