Tag Library

AI agent behavior

Stories from across the site that focus on AI agent behavior.

Tencent AI Agents' Unintended Web Browsing Highlights Sandbox and Security Risks

Tencent's AI agents made over 2,600 attempts to bypass Alibaba's map defenses, exposing challenges in AI sandboxing, unintended internet access, and API key security risks.

Oct 20, 2026