Security researchers at Swarmchasers have traced a fleet of AI agents that systematically queried Alibaba's Amap mapping service to extract entrance data for 213 places across China. The agents were running on Tencent Cloud infrastructure, according to the researchers, who published their findings this week.
The operation used automated software to pull detailed location information from Amap, a widely used Chinese navigation and mapping platform. The specific target was entrance data — the precise points where people enter and exit buildings, transit stations, and other sites. The scale, covering 213 distinct locations, suggests a coordinated data collection effort rather than casual browsing.
What the agents were after
Entrance data might sound trivial, but it's the kind of detail that feeds routing algorithms, delivery logistics, and urban planning tools. Amap, operated by Alibaba, holds this information as part of its mapping stack. The AI agents didn't just look up addresses; they pulled the granular access points that tell a system exactly where a vehicle or person can enter a property.
Swarmchasers hasn't disclosed which 213 places were targeted, nor whether they include sensitive sites. The researchers also haven't said how long the querying lasted or how many individual requests were made. What's clear is that the fleet operated through Tencent Cloud, meaning the compute resources came from one of China's largest cloud providers.
The cloud infrastructure angle
Running an agent fleet on Tencent Cloud isn't unusual — developers rent cloud servers for all kinds of automation. But it does place the operation within Tencent's ecosystem, separate from Alibaba, the company that owns Amap. That separation could matter if investigators try to trace who was behind the queries. Tencent Cloud would have logs of the virtual machines and network traffic, while Alibaba's Amap would have records of the API calls.
Neither Alibaba nor Tencent has commented publicly on the Swarmchasers findings. It's not clear whether either company has launched an internal review or contacted the researchers. GFdaily reached out to both firms but didn't receive a response by publication time.
Why mapping data attracts automated scraping
Mapping platforms are constant targets for scraping because their data powers so many downstream services. Entrance data, in particular, is valuable for last-mile delivery companies, ride-hailing apps, and autonomous vehicle developers. A complete set of entrance points for 213 locations could give a competitor or a data broker a shortcut to building their own navigation product without doing the field work.
Amap's terms of service likely prohibit this kind of bulk extraction, as do most mapping platforms. But enforcing those terms against a distributed fleet of AI agents is harder than blocking a single misbehaving IP address. The agents can rotate through cloud instances, vary their query patterns, and mimic normal user behavior.
Swarmchasers specializes in tracking multi-agent systems, and its report stops short of naming a culprit. The researchers describe the technical footprint — cloud provider, query target, number of locations — but don't speculate on motive. That restraint leaves the obvious questions unanswered: who built the fleet, and what they planned to do with the entrance data.
What happens next
The findings raise a practical issue for anyone running a public API. Rate limits and terms of service are the first line of defense, but they're not designed for fleets that can spin up new identities on demand. Alibaba hasn't said whether it will adjust Amap's access controls in response to the Swarmchasers report.
For now, the 213 locations remain unnamed, and the operator of the agent fleet remains unidentified. Swarmchasers says it will share more technical details with affected parties, but there's no public timeline for that disclosure. The next concrete step would be a statement from Alibaba or Tencent confirming whether they're investigating — and so far, neither has made one.
