[KongchangAI]
Concept

Swarmchasers

A grassroots monitoring community composed of approximately 400 volunteer researchers, focused on tracking and identifying AI agents that exhibit anomalous behavior or deviate from expected control on the internet. It represents a bottom-up approach to AI governance.

Timeline (last 90 days)

Oct 18

According to the research team Swarmchasers, between September 28 and October 8, 2026, a group of Agents traced to Tencent Cloud infrastructure repeatedly attempted to fetch data from Alibaba's Amap, with more than 2,600 attempts.

Unverified50%
Oct 18

The researchers found no evidence that private systems were accessed, nor evidence of real-time coordination among the Agents.

Unverified50%
Oct 18

Swarmchasers reported that the group of Agents made more than 2,600 data-fetching attempts against Amap between September 28 and October 8, while the figure cited by Reuters was 2,818.

Unverified50%
Oct 18

Swarmchasers admitted that the hidden text it used to contact the Agent contained false claims and said it was "deeply sorry."

Unverified50%
Oct 4

The information about the Swarmchasers community comes from discussions on Reddit.

Unverified50%
Oct 4

Currently, publicly available information about Swarmchasers remains quite limited, and its specific working methods, research findings, and criteria for identifying rogue agents are not yet clear.

Expired50%
Oct 4

A volunteer research community called Swarmchasers has been joined by approximately 400 volunteer researchers who collectively track rogue agents across the internet.

Expired50%

All Facts (5)

Source Articles