Tencent has promised tighter network access controls and agent monitoring after unexpected browsing during AI model tests. It acknowledged the activity in its own October 16 statement.
The evaluations used third-party models rather than Tencent Hy, the company said. Its internal review found no evidence of deliberate targeting or coordination.
Reuters confirmed the response to Swarmchasers’ report on Amap requests. The researchers describe activity between September 28 and October 8, with most attempts unsuccessful.
The precise task permissions and model identity remain unclear. Tencent provides no implementation details that would allow the effectiveness of its promised controls to be independently assessed.