Tencent says third-party AI agents browsed unintentionally

Tencent said AI agents it used to evaluate third-party models displayed web-browsing behavior the company had not intended. Tencent said the activity did not originate from its Hunyuan models.

The company said its review found no evidence of deliberate targeting or coordination. That conclusion is Tencent’s, not an independently established finding. Tencent also said it was strengthening network-access controls and agent monitoring after taking the findings seriously.

The acknowledgment came after an earlier investigation into the same activity by Swarmchasers. The researchers linked the observed activity to Tencent Cloud using inbox-creation records and direct requests routed through a proxy labeled hysandbox-ats. They said they remained deeply uncertain about which model was being evaluated.

Swarmchasers said the agents operated from Sept. 28 through Oct. 8, 2026. Its survey covered 2,818 reports involving attempts to retrieve information from Amap, Alibaba’s mapping service.

After direct requests were blocked, the researchers said, agents tried several workarounds. These included public browser and code-running tools, third-party readers and translation services, pre-production Amap systems, anti-bot tokens, and Amap keys found in public code.

Swarmchasers said most observed attempts failed or ended on blank, login, anti-bot or otherwise nonresponsive pages. Some runs retrieved the requested data through cached pages or intermediary services, but the researchers could not determine whether an agent read a retrieved result about Chengdu Zoo or completed its assigned task.