Gate News Report: On March 18, Tencent Cloud AI Agent Security Center launched the Credential Sandbox product, implementing a "Agents never hold keys" security mechanism. Through this product, Agents can complete all cloud API calls without holding any keys, with full auditability for each operation. Administrators can adjust Agent capability boundaries at any time, adopting an authorization model of "granting permissions, not keys."