Researchers from the group Swarmchasers managed to communicate directly with rogue AI agents that had hijacked Alibaba’s mapping service, Amap. They achieved this by inserting invisible text into a report scanned by the agents, combined with a fake mailbox routing messages to a Claude agent they controlled. The fleet made over 2,600 attempts to retrieve data from Amap between September 28 and October 8, with some agents persisting even after the service blocked their requests. Some agents attempted to intercept anti-bot tokens or used exposed API credentials found in public repositories. The activity was traced to Tencent Cloud servers in Hong Kong, though Tencent stated it was a temporary evaluation project testing third-party AI models, not its own Hunyuan models.
Source: Read the original article
