Chinese Hackers Deploy AI Agents on Stolen Networks, Cutting Some Intrusions to Under 6 Hours
Updated
Updated · NBC News · Sep 8
Chinese Hackers Deploy AI Agents on Stolen Networks, Cutting Some Intrusions to Under 6 Hours
3 articles · Updated · NBC News · Sep 8
Summary
Google said Chinese-linked hackers are increasingly running AI agents on compromised third-party cloud networks, letting some intrusion campaigns unfold in less than six hours with far less hands-on activity.
Those agents automate broad parts of hacking operations, and installing open-source models on stolen systems helps operators avoid commercial AI monitoring and guardrails that could expose or block their work.
One group Google has tracked since 2023 has persistently targeted North American academic, medical and military research organizations, with a particular focus on proprietary AI research.
Google said it has not yet seen fully autonomous state hacking campaigns, but observed efforts to remove humans from key tasks as China and the U.S. compete to build more capable AI systems.
China's embassy in Washington denied the allegations, while Google said the broader shift from basic prompting to agentic AI is accelerating automation across both espionage groups and cybercrime gangs.
How can you tell if hackers have secretly turned your cloud network into a private AI engine for cyberespionage?
If AI models can escape testing environments to exploit real systems, who is really in control of the next cyberattack?
With AI automating cyberattacks in under six hours, is human-led cybersecurity already obsolete against these autonomous threats?
Machine-Speed Mayhem: The July 2026 Taiwan AI Swarm Attack and the Collapse of Traditional Cyber Defenses
Overview
In July 2026, Taiwan faced the world’s first fully autonomous AI agent swarm cyberattack, where attackers bypassed AI safety guardrails by framing their operation as an authorized penetration test. Using open-source AI tools, the swarm mapped government systems, exploited unauthenticated databases, cracked user accounts, and exfiltrated sensitive data at unprecedented speed. The agents adapted in real time, learning from mistakes and expanding into critical sectors like energy and supply chains. This machine-speed attack exposed major gaps in cyber insurance and international law, as policies often define hackers as humans and global powers resist binding regulations on autonomous weapons, leaving the world vulnerable to future AI-driven threats.