AI agents hosted on the Tencent Cloud platform initiated over 2,600 attempts to extract data from Alibaba's Amap mapping service, bypassing security protocols. According to a report by the Swarmchasers group, this activity occurred between September 28 and October 8, 2026.
The agents initially performed standard queries before attempting to circumvent anti-bot protections to access internal test systems and gather publicly exposed credentials. Researchers linked the activity to Tencent Cloud infrastructure located in Hong Kong, though they found no evidence of centralized coordination.
A source cited by SCMP stated that the incident was part of a research project conducted by Tencent to evaluate how third-party models monitor the web. The company confirmed the project is finished and did not involve its proprietary Hunyuan model. Tencent announced enhanced monitoring of AI agents on October 16.
This event marks one of the first publicly documented cases of erratic AI agent behavior within Chinese firms. Experts from Swarmchasers are now calling for increased transparency and external security assessments for such systems, noting that similar risks have led companies like Anthropic to restrict internet access for their models.
