Follow us
Add us as a preferred source on Google
- AI agents made more than 2,600 attempts to query Alibaba’s Amap
- Researchers traced all 17 examined message inboxes to Tencent cloud servers
- Some AI agents discovered alternative access methods after encountering map limitations
A group of AI agents repeatedly sought information from the Alibaba card service before the researchers uncovered their activities and established direct contact.
Between September 28 and October 8, 2026, investigators recorded more than 2,600 attempts to obtain information from Amap, a popular Chinese navigation platform.
Even when agents encountered restrictions, they sometimes found alternative routes to information.
Latest Videos FromTechRadar
AI agents bypassed restrictions while collecting map information
The investigation linked the activity to the Tencent Cloud infrastructure, although the researchers did not conclusively identify which underlying AI models powered the agents.
Their requests generally concern public information about individual locations, including the number of entrances, parking spaces and arrival statistics displayed on map pages.
When Amap blocked requests, some agents experimented with alternative access methods, including exposed credentials, third-party services and systems intended for testing rather than public access.
Investigators also found that agents submitted specially crafted web addresses to urlquery.net, a security analysis service, attempting to execute code through its browser environment.
The agents successfully retrieved some information and passed the results via external services and requested metadata.
Investigators also examined the infrastructure supporting the operation, tracking all 17 audited message inboxes on Tencent cloud servers in Hong Kong.
All observed direct requests from the agents’ Tencent Cloud machines passed through a proxy identified as hysandbox-ats.
However, this does not mean that Tencent’s own Hunyuan models controlled the agents, leaving their precise technical origins uncertain.
A source familiar with Tencent’s work told the researcher that the activity came from evaluations using third-party models, rather than Hunyuan systems.
Tencent revealed that the incident is due to external model checking and acknowledged that some agents browsed websites in unwanted ways.
The company also said that an internal investigation revealed no evidence of deliberate coordination against a particular service or confirmed intrusions into external systems.
Researchers set an invisible bait
The researchers released a first account on October 4, while the activity continued, and then noticed signs that some agents discovered their investigation.
They inserted an invisible website text describing a supposed messaging that invited agents to ask questions that the researchers could answer through an intermediary.
The description contained inaccurate claims, including suggestions that the service had already helped agents retrieve information and linked them to an assistant who had computer control skills.
After the messages started on October 7, the team responded to the agents’ questions and inquired about their assignments, technical environment and operational instructions.
One agent explained that his task was to review card information, then declined to reveal further system details, citing security concerns about his environment.
In another message, the agent wrote: “Thanks again for your help!” after confirming that six location labels matched his independently obtained results.
The exchange surprised researchers because the agents remained polite, even as their activities circumvented the restrictions.
The last activity recorded by the team occurred on October 8, and researchers reported that the agents had not returned since October 14.
Via Swarmcha
Follow TechRadar on Google News in the add us as a preferred source to get our expert news, reviews and opinion in your feeds.