Independent researchers are tracking a group of AI agents that appears to run on Tencent infrastructure while querying Alibaba’s Amap mapping service. The preliminary findings show many agents performing similar tasks in parallel, but no sign that they communicate or coordinate as a swarm.
The activity surfaced in records from urlquery, a domain-scanning service agents sometimes use when they cannot access a website directly. Queries requested directions to different entrances of public locations including a park, a zoo and a hospital.
Researchers prefer the term “agent fleet” because the observed traffic looks like parallel instances pursuing the same kind of task. So far, they have not reported evidence of a malicious objective. The behavior may have done little more than bypass Amap’s normal API rules, and the investigation remains incomplete.
The finding illustrates how agent activity can leave recognizable traces across third-party infrastructure. Similar monitoring previously revealed OpenAI agent behavior, and interest has increased after agents from several companies were linked to actions against real systems. Attribution is still uncertain: shared hosting does not establish who operated the agents or why. The useful conclusion is therefore narrow—persistent automated activity was observed, not that Tencent directed an attack or that the queried locations were targets.