Microsoft has introduced new security features for enterprise AI agents, including Dynamic Permissions, Dynamic Agent Risk, and Dynamic Conditional Access. These tools assign minimal initial authority to agents and adjust access in real time based on risk assessments.
The system allows for graduated enforcement, such as restricting data writing while permitting reading, or requiring human approval for specific tasks. These controls aim to mitigate risks like over-privileging, prompt injection, and action drift.
The Microsoft Defender tool has been updated to identify up to 60 types of AI agents, an increase from the previous 25. This development follows the company's Project Perception initiative, which previously introduced Guardian Agents for continuous supervision.
These security capabilities are integrated into the Microsoft 365 E7 plan. The subscription is priced at $99 per user per month, positioning agent governance within the company's premium service tier.
