Researchers Trace Chinese AI Agent Swarm Running on Tencent Infrastructure, Targeting Alibaba's Amap

Independent security researchers have identified a coordinated network of AI agents that appears to be hosted on Tencent's cloud platform and is directing activity at Amap, the mapping service owned by rival Alibaba. The finding, reported by TechCrunch, points to one of the first publicly documented cases of an agent fleet deployed at scale across competing Chinese tech infrastructure.

The swarm consists of multiple autonomous software agents that operate in concert, according to the researchers, who tracked the traffic and linked the agents to Tencent's servers. Their target, Amap, is one of China's most widely used navigation and location services, with hundreds of millions of monthly users. The agents' purpose—whether data scraping, load testing, or something else—remains unclear from the available evidence.

Agent swarms differ from traditional botnets in that each agent can make independent decisions, communicate with peers, and adapt its behavior based on feedback. That makes them harder to detect and block than scripted traffic. Researchers have warned for months that such systems could be used for automated reconnaissance, competitive intelligence, or denial-of-service campaigns, though concrete examples have been scarce until now.

The involvement of Tencent infrastructure adds a layer of complexity. Tencent and Alibaba are direct competitors in cloud computing, payments, and consumer internet services, but there is no indication that Tencent itself is behind the activity. Cloud platforms routinely host third-party workloads, and the researchers have not attributed the swarm to any specific operator. Tencent has not commented publicly on the finding.

This discovery lands as AI agents move from research demos into production tools. Companies including OpenAI, Anthropic, and several Chinese firms have released frameworks that let multiple agents collaborate on tasks, and businesses are beginning to deploy them for customer service, coding, and data analysis. The same capabilities that make agent fleets useful also make them suitable for hostile automation, a dual-use problem that security teams are only starting to confront.

What happens next depends largely on whether the researchers can identify who controls the swarm and what it is doing. Amap's parent Alibaba has not issued a statement, and it is not clear whether the activity has caused any service disruption. For now, the episode serves as a live case study in how autonomous agents behave when they escape the lab—and how difficult they are to trace once they do.