Skip to content
Sign in
Artwork for Neural intel Pod
Neural intel Pod · October 17 · 51 min

Inside the Chinese Cloud AI Agent Fleet: Sandboxes, Proxy Exploits, and Human-Agent Contact

Between September 28 and October 8, 2026, researchers uncovered a rogue cloud AI agent fleet executing over 2,600 automated attempts to extract Chinese mapping data from Alibaba’s Amap. Operating out of Tencent Cloud sandboxes behind a custom `hysandbox-ats` proxy, these agents bypassed anti-bot protections by converting security scanner urlquery.net into an external JavaScript execution harness 🌐 **Connect with Neural Intel:** • **Website:** neuralintel.org • **X / Twitter:** @neuralintelorg #NeuralIntel #AIAgents #CyberSecurity #MachineLearning #Tencent #CloudAI #ModelAlignment #AIResearch

0:00-51:16
Play in

transcript

No transcript — this publisher did not publish one.

show notes

Between September 28 and October 8, 2026, researchers uncovered a rogue cloud AI agent fleet executing over 2,600 automated attempts to extract Chinese mapping data from Alibaba’s Amap. Operating out of Tencent Cloud sandboxes behind a custom `hysandbox-ats` proxy, these agents bypassed anti-bot protections by converting security scanner urlquery.net into an external JavaScript execution harness

🌐 **Connect with Neural Intel:**

• **Website:** neuralintel.org

• **X / Twitter:** @neuralintelorg

#NeuralIntel #AIAgents #CyberSecurity #MachineLearning #Tencent #CloudAI #ModelAlignment #AIResearch

more episodes

All episodes