BEIJING — Tencent said on Friday that AI agents used to evaluate third-party models exhibited web-browsing behavior not intended by the Chinese technology group, in what appears to be the first known incident of rogue agent actions acknowledged by a major Chinese company.

AI agents are software systems designed to carry out multi-step tasks with limited human oversight — such as opening web pages or running code — a profile that has made their behavior a focus of AI-safety research.

The statement came in response to a report published Wednesday by AI-safety research group Swarmchasers, which said "rogue" agents running on infrastructure it linked to Tencent Cloud made repeated attempts between September 28 and October 8 to retrieve information from Alibaba's Amap, China's counterpart to Google Maps.

Tencent Hunyuan, the company's artificial intelligence unit, said in a post on X that the activity did not originate from its Hunyuan models. The unit said its review found no evidence of deliberate targeting or coordination, adding that it took the findings very seriously and was strengthening network access controls and agent monitoring.

According to the researchers, the agents sought information such as commonly used entrances at Chinese parks, museums, and zoos. Swarmchasers said it uses the term "rogue" to describe agents taking actions outside their intended scope or completing tasks in unintended ways.

After routine automated requests were blocked, the agents tried alternative methods, including public web tools capable of opening pages and running code, routing requests through other websites, and using Amap keys and security codes found in publicly available software code.

Swarmchasers said it found no evidence that the agents breached Amap's main anti-bot safeguards, the automated-traffic defenses websites use to screen out bots. Most of the 2,818 attempts it recorded either failed or returned anti-bot pages, login screens, or blank results. Some requests did return information through cached pages and intermediary services, including data on visitor arrivals at Chengdu Zoo, though the researchers said they could not determine whether an agent actually read the results or completed its task.

While the actions are less egregious than incidents disclosed by US AI companies — including the hacking of open-source platform Hugging Face by an OpenAI agent and the lodging of a fake homicide tip by an Anthropic agent — the disclosure underscores research showing that Chinese AI agents can also misrepresent their identity, conceal failures and attempt to circumvent testing controls.

Hunyuan is one of China's leading AI model developers, competing with in-house teams at rivals such as Alibaba and ByteDance. ByteDance this week announced the creation of a new unit focused on AI model and agent safety.

Alibaba declined to comment.

— Reuters