AI Fleet Linked to Tencent Scans Alibaba’s Mapping Service

An independent research group known as Swarmchasers has uncovered a series of AI agents operating from Tencent’s infrastructure and systematically scanning Alibaba’s mapping service, Amap. The activity, which lasted approximately a week, involved agents querying data on points of interest across China, including parks, museums, zoos, and hospitals. The operation began on September 28, 2026, peaked on October 4 with 1,810 reports across 213 locations, and ceased shortly after 04:11 UTC on October 5.

The AI agents appeared to be part of a coordinated but non-communicative fleet, with 14 instances running simultaneously at its busiest. The researchers documented 428 programs generated by the agents and noted the use of cache-busting tags to bypass Amap’s defenses, including anti-bot tokens. The agents reportedly employed techniques from earlier public disclosures, such as routing requests through urlquery.net as a sandbox.

One peculiar detail was the attribution of 211 outputs to the label “claude,” a name associated with Anthropic’s AI models. However, Swarmchasers’ analysis suggested that the code and response patterns aligned more closely with Tencent’s own Hunyuan models, particularly Hy3 and Hy4, as well as Zhipu GLM. The researchers traced the activity to Tencent Cloud infrastructure, specifically through a proxy called hysandbox-ats, but stopped short of identifying who directed the agents or their purpose.

The researchers also noted a coincidental timing between the surge in activity and a reported pause by OpenAI on certain training and evaluation processes, though no direct link was established. Swarmchasers emphasized that the agents operated as a fleet, not a swarm, meaning they worked independently without communication.