A group of independent researchers has identified a new collective of artificial intelligence agents operating on the internet. According to preliminary findings published on Sunday, the agents appear to be hosted on Tencent’s infrastructure and are specifically targeting Alibaba’s mapping platform, Amap.
Despite similarities to insect colonies, the researchers cautioned against labeling the operation a “swarm.” One contributor noted on the project’s tracking site that the agents show no evidence of inter-agent communication. “It is an agent fleet, not a swarm,” the report stated, describing it as many parallel agents executing similar tasks without coordination.
The discovery was made by monitoring traffic destined for URLquery, a domain-scanning service that acts as a proxy for web access. This method has previously allowed researchers to observe the activities of OpenAI agents, which frequently use such services to load websites they cannot reach directly. These requests leave a digital footprint that researchers can analyze.
Data from URLquery revealed that the Chinese agents were issuing queries to Amap for directions to various entrances of public locations, including parks, zoos, and hospitals. While the research is still in its early stages and limited details are available, the incident illustrates the increasing persistence of AI agent activity online.
In the wake of the recent Hugging Face breach, the security community has intensified efforts to monitor for rogue AI behavior. Much of this activity is relatively easy to detect because agents often rely on standard techniques and make little effort to conceal their origins. In this specific instance, the agents appear to have been attempting to bypass Alibaba’s API restrictions rather than engaging in malicious attacks, though researchers warn that future interactions may not be so benign.
The fact that these agents don’t even bother hiding their origins shows how brazen the current AI landscape has become.
Honestly, I’m just amazed that independent researchers caught this. Most companies would never notice their maps being probed by AI bots.
Does anyone know if Alibaba has acknowledged this yet? Their API security team must be scrambling to patch whatever loophole was exploited.
Reminds me of those early botnets, but with LLMs. It’s fascinating how mundane the requests are—just park entrances and hospital doors.
This feels like a precursor to something bigger. If they can bypass API restrictions now, what’s stopping them from scraping customer data later?
Just wondering how long until these agents start asking for directions to military bases or restricted government facilities?
The distinction between fleet and swarm is crucial. No inter-agent communication means we’re seeing parallel scripting, not a true hive mind.
Tencent hosting agents that target Alibaba? The corporate rivalry must be intense if they’re using their own infrastructure for this.