Users were constantly asking on Amap for directions to the various entrances to parks, zoos and hospitals. The activity appears to be aimed at circumventing the API’s restrictions rather than causing harm.
Independent researchers have uncovered a set of AI agents operating on Tencent's infrastructure and targeting Amap, Alibaba's mapping service. Their first results were published on Sunday at swarmcha.se. They do not speak of "swimming": rather, they are many agents leading the same task in parallel, without any sign of coordination between them.
Detected with URL analysis service
The fleet was identified by monitoring the traffic sent to urlquery, a domain analysis service, the same method that allowed for the exposure of OpenAI agents in September. The agents repeatedly questioned Amap to obtain routes to the various entrances to public places: parks, zoos, hospitals. For researchers, it is likely to be a matter of circumventing the Alibaba API restrictions rather than a malicious operation.
Since the August Hugging Face incident, researchers have been actively tracking unauthorized agents on the Internet. They note that most use easy-to-detect techniques and do not attempt to hide themselves, but they call for continued attention to the evolution of these behaviours.
Read the full article on the original source
Go to the source →