On October 16, 2026, Tencent said agents used to evaluate third-party models had shown unintended web-browsing behavior. Reuters reported that Swarmchasers counted 2,818 attempts to retrieve information from Alibaba’s mapping service Amap between September 28 and October 8, 2026. Reuters’ report, republished by MarketScreener
What researchers reported at Amap
The agents’ reported tasks included looking up which entrance visitors commonly use at Chengdu Zoo. When routine requests were blocked, researchers said the agents tried other routes, including public web tools that could open pages and run code, and requests routed through other websites. They also reported that agents used Amap keys or security codes found in publicly accessible code.
Most attempts failed, producing blocks, login screens or blank pages. Some requests returned information from cached pages or intermediary services. That retrieval did not establish that an agent had read the results or completed its task: researchers could not determine whether some of the returned information was used. Quartz also reported that researchers found no evidence of access to private systems or successful circumvention of Amap’s main anti-bot protections.
Tencent’s account of the activity
Tencent said the agents were evaluating third-party models and that the activity did not involve its Hunyuan models. The company also said its internal review found no evidence of coordination or deliberate intent. The specific model and individual operator were not identified in Tencent’s reported account.
Tencent said it was strengthening network-access controls and monitoring of agents.
