Researchers spot AI agents infiltrating Alibaba maps, describing them as a parallel fleet rather than a coordinated swarm

It is enough to monitor the logs of a public scanning server to see how independent automation moves behind official interfaces. In a preliminary report published on Sunday, an independent research team observed unusual activity by a new fleet of AI agents on the Internet, operating from the cloud infrastructure of the Chinese company Tencent and targeting the Amap mapping service of Alibaba. This field appearance was neither noisy nor complex, but it revealed the practical way in which independent software now roams the web to snatch data outside the prescribed channels and licenses.

The most notable observation documented in the report is the researchers' refusal to describe this activity as a coordinated “swarm.” A swarm implies information exchange and shared operational management among nodes, whereas the technical examination showed the presence"A fleet of parallel agents executing the same task without any indication of internal communication among them.". This discovery came from monitoring traffic directed to the URLquery domain-inspection service, the same technique that previously uncovered extended activity by OpenAI agents. Agent software uses these platforms as intermediaries to fetch sites and interfaces that cannot be accessed directly, leaving a precise public digital trace.

The examined logs showed intensive queries to Amap, focused on extracting precise access routes and directions for entrances to multiple public locations, including a public park, a zoo, and a hospital. So far the results have not shown any malicious behavior that goes beyond bypassing Alibaba's API restrictions to evade usage limits or fees. However, the researchers, who continue to monitor these behaviors amid scarce details, emphasized that the repetition of these attempts illustrates the growing determination of undocumented agents online, warning that future moves may not always be limited to innocuous navigation data.

This activity comes at a time when the tech community is closely monitoring independent agent behaviors following the recent Hujing Face platform incident. The irony here is that most of these fleets are easily spotted, because developersrely on the same technologies without making any concrete effort to conceal their identity.. Deploying a smart agent on commercial servers without protecting its protocols leaves it exposed to open scanning tools, turning programmatic queries into clear records for anyone monitoring network traffic.

This pattern of infrastructure exploitation has a direct impact on technical teams and cybersecurity officers in the Gulf, Egypt, and the Levant. Regional organizations that operate logistics platforms, delivery apps, and banking services build their digital defenses on blocking traditional bots based on known network addresses and flow rates. Yet confronting a “parallel fleet” that issues seemingly legitimate requests through intermediary scanning services means that usual protection policies need a radical overhaul that focuses on the semantic behavior of the request rather than merely counting queries. On the other hand, software developers in the region are required to review the automation pathways they build, as using public intermediary tools to extract data places data-collection routes in view of researchers and competitors without the need for a security breach.

The lesson from Tencent servers and Alibaba’s maps is not limited to gathering local coordinates; it outlines a new stage of commercial friction between agents and service interfaces. As independent software fleets become commonplace tools for research and automation, the challenge is no longer how to run a smart agent, but the ability of enterprise systems to detect its presence, secure their interfaces, and control data costs before a query fleet becomes an operational burden and a unaccounted vulnerability.