THR857 Theater Thu 19 Nov 17:30 PST
Secure, Govern, and Scale: Protecting AI Agents with Microsoft Agent 365
Brendan Powers
Summary
From the recording
Brendan Powers demonstrated Agent 365 in a live tenant, showing an action centre listing 160 high-risk agents, a dynamic agent risk score built on runtime, exposure and blast radius risk, and a third-party ServiceNow agent with an Entra agent ID. Sierra and Darktrace feed in signals. Agents can be onboarded through Copilot Studio auto-registration, the SDK or connected platforms. Ingesting 1,000 Anthropic Claude agents gives cost visibility, security value and observability. Caveat: no GA, preview, pricing or licensing details were stated; some features were described only as coming.
Session description
Microsoft's catalogue description
See how Microsoft Agent 365 brings agent context and security evidence together. Follow demonstrations of registered-agent security and Microsoft Defender investigation, with Cyera and Darktrace perspectives on data and behavioral signals. Explore how proposed partner integrations could inform agent-risk assessment, then learn to choose supported responses. Leave with a practical workflow to discover, assess, investigate, and protect AI agents.
Demo timestamps
Microsoft's player does not support links to a specific time, so open the recording and skip to the time shown.
- Skip to 1:26
Protecting and governing agents with Agent 365
First walkthrough on protecting and governing AI agents with Agent 365, introduced in the captions. Later demo transitions occur at 8:47 and 16:45.