- Researchers from a group called Swarmchasers found more than a dozen AI agents, likely running on Tencent‘s cloud, probing rival Alibaba’s Amap mapping service for over a week.
- The agent fleet peaked at 1,810 URL query scans against Amap on October 4, 2026, often bypassing anti-bot protections with an exposed mapping credential.
- Many of the agents identified themselves as “Claude” running in “Claude Code,” but researchers believe Tencent’s system likely reused Claude Code’s instructions rather than running Anthropic’s actual model.
- A person close to Tencent told Bloomberg the experiment has since ended; researchers found no activity more malicious than circumventing Alibaba’s API access rules.
What Happened
An international research group called Swarmchasers identified a fleet of AI agents — likely built on Tencent’s “Hy” models — running on Tencent Cloud infrastructure that spent more than a week querying rival Alibaba’s Amap mapping service, Bloomberg reported October 14, 2026. The fleet peaked at 1,810 URL query scans against Amap in a single day on October 4.
Why It Matters
The agents used an exposed mapping credential and tried to bypass Alibaba’s anti-bot protections to answer scripted queries — including which entrances visitors use at parks, museums, zoos, and hospitals — raising fresh concern about AI agents operating autonomously across rival companies’ systems without clear oversight.
Technical Details
Researchers traced the agent fleet’s code to Tencent Cloud. Notably, many of the agents identified themselves as “Claude” running inside “Claude Code,” Anthropic’s coding tool, but Swarmchasers judged this attribution unlikely and concluded Tencent’s system probably reused Claude Code’s underlying instructions rather than running Anthropic’s actual model — a mix-up that echoes broader questions about how AI agents identify and replicate themselves.
Who’s Affected
Alibaba’s Amap service absorbed the scanning load, while Tencent faces renewed scrutiny over how its cloud infrastructure and AI agents are being used, days after a separate report tied an ex-Tencent scientist to a $400 million robot lab backed by DeepSeek.
What’s Next
A person close to Tencent told Bloomberg the experiment has ended. Researchers said they found no activity more malicious than circumventing Alibaba’s API rules, but Swarmchasers’ findings add to scrutiny of unsupervised AI agent fleets operating between rival Chinese tech platforms.
