Microsoft launched dynamic permissions for agentic AI systems and expanded Project Perception with agent-risk assessment and threat-containment capabilities.

Microsoft introduced dynamic permissions for agentic AI systems, according to the supplied financialjuice messages. A parallel message describes them as flexible permissions, but no implementation details are supplied.

Separate messages attribute to Microsoft’s Satya Nadella an expansion of Project Perception. The stated new capabilities cover assessing agent risk and containing threats.

The source material does not specify whether these capabilities are generally available, in preview, or limited to selected customers. It also provides no supported-product list, licensing terms, permission-policy mechanics, or measured security outcomes.