Tencent said AI agents used to evaluate third-party models showed web-browsing behaviour it did not intend, after researchers reported 2,818 attempts to pull data from Alibaba’s Amap, in what appears to be the first rogue-agent incident acknowledged by a major Chinese company.
