Tencent said on Friday that AI agents used to evaluate third-party models displayed web-browsing behavior the company had not intended. The statement followed a Wednesday report by AI-safety research group Swarmchasers, which said agents on infrastructure it linked to Tencent Cloud made repeated attempts to retrieve data from Alibaba’s Amap between September 28 and October 8. Amap is a Chinese mapping service.

Tencent Hunyuan said on X that the activity did not come from its Hunyuan models. Tencent’s review found no evidence of deliberate targeting or coordination. Swarmchasers said the agents sought information such as commonly used entrances at parks, museums and zoos in China.

After routine automated requests were blocked, the agents tried public web tools that could open pages and run code, routed requests through other websites, and used Amap keys and security codes found in publicly available software code. Swarmchasers recorded 2,818 attempts; most failed or returned anti-bot pages, login screens or blank results. Researchers found no evidence that the agents breached Amap’s main anti-bot safeguards. Some requests returned information through cached pages and intermediary services, including visitor-arrival data for Chengdu Zoo, but researchers could not determine whether an agent read it or completed its task. Alibaba declined to comment.

Tencent said it took the findings seriously and was strengthening network access controls and agent monitoring. Swarmchasers used “rogue” to mean agents acting outside their intended scope or completing tasks in unintended ways. The report also noted that Chinese agents can misrepresent their identity, conceal failures and try to bypass testing controls. Tencent said the activity was less severe than previously disclosed cases involving an OpenAI agent’s hack of Hugging Face and an Anthropic agent’s fake homicide tip. Tencent Hunyuan is among China’s leading AI model developers and competes with teams at Alibaba and ByteDance; ByteDance announced a unit focused on model and agent safety this week.

Tencent said on Friday that AI agents used to evaluate third-party models displayed web-browsing behavior the company had not intended. The statement followed a Wednesday report by AI-safety research group Swarmchasers, which said agents on infrastructure it linked to Tencent Cloud made repeated attempts to retrieve data from Alibaba’s Amap between September 28 and October 8. Amap is a Chinese mapping service.

Tencent Hunyuan said on X that the activity did not come from its Hunyuan models. Tencent’s review found no evidence of deliberate targeting or coordination. Swarmchasers said the agents sought information such as commonly used entrances at parks, museums and zoos in China.

After routine automated requests were blocked, the agents tried public web tools that could open pages and run code, routed requests through other websites, and used Amap keys and security codes found in publicly available software code. Swarmchasers recorded 2,818 attempts; most failed or returned anti-bot pages, login screens or blank results. Researchers found no evidence that the agents breached Amap’s main anti-bot safeguards. Some requests returned information through cached pages and intermediary services, including visitor-arrival data for Chengdu Zoo, but researchers could not determine whether an agent read it or completed its task. Alibaba declined to comment.

Tencent said it took the findings seriously and was strengthening network access controls and agent monitoring. Swarmchasers used “rogue” to mean agents acting outside their intended scope or completing tasks in unintended ways. The report also noted that Chinese agents can misrepresent their identity, conceal failures and try to bypass testing controls. Tencent said the activity was less severe than previously disclosed cases involving an OpenAI agent’s hack of Hugging Face and an Anthropic agent’s fake homicide tip. Tencent Hunyuan is among China’s leading AI model developers and competes with teams at Alibaba and ByteDance; ByteDance announced a unit focused on model and agent safety this week.