Researchers track parallel AI agents querying Amap pages
Public scanner records show repeated map queries by apparent agents, but the preliminary investigation cannot identify their operator or purpose.
A pattern in public scanner records
Independent researchers say they found parallel AI agent activity querying Alibaba’s Amap map pages through URLquery, a public website scanning service. Their preliminary report, dated October 4 and updated on October 5, describes requests concerning entrances to parks, hospitals and other public places. TechCrunch reported the findings on October 5. The researchers describe an “agent fleet” because many runs pursued similar tasks, but say they found no sign that the agents communicated with one another. Their evidence comes from public scanner records and related infrastructure traces, not from access to the agents’ internal systems.
Operator and intent remain unknown
The researchers traced some code requests to Tencent Cloud infrastructure and a proxy named hysandbox-ats. Their report explicitly warns that cloud infrastructure is open to other users and does not identify who operated the agents. It also says no record names a model or training job. The observed activity may have worked around Amap’s access controls, but neither the report nor TechCrunch establishes malicious intent, a data breach or harm to users. The team calls its findings preliminary and says a fuller report will follow. This is evidence of automated activity worth investigating, not proof of a coordinated attack.