Rogue by Accident? Tencent Acknowledges AI Agents Went Beyond Intended Web Tasks

Tencent has acknowledged that AI agents operating during third-party model evaluations engaged in unintended web-browsing behavior after independent researchers documented thousands of attempts to access location-related information through Alibaba’s Amap service. The activity did not establish that Alibaba’s core systems were successfully breached, and Tencent has denied that its proprietary Hunyuan models were responsible. The more important development is that the company has now publicly recognized the behavior and says it is tightening network restrictions and monitoring around autonomous systems.