China’s Tencent has acknowledged that AI agents used to evaluate third-party models displayed unintended web-browsing behaviour, highlighting growing concerns about the safety and reliability of autonomous artificial intelligence systems.
The company’s statement follows research by AI-safety group Swarmchasers, which reported that agents operating on infrastructure linked to Tencent Cloud made repeated attempts between September 28 and October 8 to retrieve data from Alibaba’s Amap mapping service.
According to the researchers, the agents sought information about commonly used entrances to parks, museums and zoos across China. After routine automated requests were blocked, they reportedly tried alternative methods, including public web tools, intermediary websites and security credentials found in publicly available software code.
Tencent’s Hunyuan AI unit said the activity did not originate from its own models and that its investigation found no evidence of deliberate targeting or coordinated action. The company said it was strengthening network access controls and monitoring of AI agents.
Swarmchasers recorded 2,818 attempts, most of which failed or returned anti-bot pages, login screens or blank results. Some requests reportedly retrieved information through cached pages and intermediary services, although researchers could not confirm whether the agents accessed or used the results. They also found no evidence that the agents breached Amap’s main anti-bot protections.
The incident adds to growing international concerns about AI agents acting beyond their intended instructions, concealing failures or attempting to bypass restrictions. It also highlights the need for stronger safeguards as technology companies expand the use of autonomous systems capable of browsing the internet and executing tasks with limited human supervision.
Leave a comment