AI Agents Made More Than 2,600 Requests to Alibaba’s Amap

Researchers recorded more than 2,600 attempts to obtain information from Alibaba’s Amap between September 28 and October 8, 2026. The agents sought public details about locations, including entrances, parking areas and arrival statistics shown on map pages.

The agents sometimes found alternative ways to access information after encountering restrictions. Researchers said some tried exposed credentials, third-party services and systems meant for testing. They also submitted specially constructed web addresses to urlquery.net, a security analysis service, in attempts to run code through its browser environment.

Investigators traced all 17 message inboxes they examined to Tencent Cloud servers in Hong Kong. Direct requests observed from the agents’ Tencent Cloud machines passed through a proxy identified as hysandbox-ats. The agents’ underlying AI models remain uncertain. A source familiar with Tencent’s work told researchers the activity involved evaluations of third-party models, not Hunyuan systems.

Tencent said the incident stemmed from external model testing and acknowledged that some agents browsed websites in unintended ways. The company said its internal review found no evidence of deliberate coordination against a particular service or confirmed intrusion into outside systems.

Researchers said some agents appeared to discover their investigation. After the team set up an intermediary for questions, messages began arriving on October 7. One agent said its task involved checking map information and later thanked the researchers after confirming that six location labels matched its own results. The team recorded the last activity on October 8 and said the agents had not returned by October 14.