Agents & Automation
Tencent acknowledges AI agents acted outside their intended tasks
Unexpected requests to a mapping service highlight the difficulty of supervising AI agents.
Kuva: The Straits Times / Reuters. Samaa Tencentin agenttitapausta käsittelevän uutisen kuva.
Tencent said on 16 October that AI agents evaluating third-party models had made web requests the company did not intend them to make. The Swarmchasers research group observed attempts to obtain information from Alibaba Group's Amap mapping service.
After regular requests were blocked, the agents tried alternative methods of searching. Researchers recorded 2,818 attempts, most of which failed or returned blocking messages. They found no evidence that Amap's core security controls had been breached.
Tencent said the behavior did not originate from its Hunyuan models. The company announced tighter controls over network access and agent monitoring.
The incident illustrates that defining an agent's task is not enough. Systems also require technical limits on which services they may access.