Skip to content

THR877 Theater Tue 17 Nov 16:05 PST

AI agents: access granted. Trust not included

Ehud Itshaki, Leandro Iwase

Summary

From the recording

Microsoft presented giving AI agents their own Entra identities rather than sharing user access. Entra Agent ID for local agents is in preview, with Defender and Global Secure Access discovering 60 plus local agents. Unified dynamic agent risk across Entra, Defender and Purview is in public preview. Conditional Access dynamic controls are in limited or private preview, while Conditional Access for agents is generally available. A demo showed risk escalating and agent access being restricted, then blocked. Licensing, pricing and dates were not stated.

Session description

Microsoft's catalogue description

AI agents need access to apps, data, and resources, but should they get a blank check? See how Microsoft Entra Agent ID helps secure agent connections, limit risky permissions, and apply network controls to reduce risk and block compromised access before it becomes an attack path.

Demo timestamps

Microsoft's player does not support links to a specific time, so open the recording and skip to the time shown.

  1. Skip to 11:09

    Agent discovery, risk scoring and visibility into local AI agents

    Live demo of discovering local AI agents, assigning risk scores and gaining visibility into them.

  2. Skip to 13:50

    Enforcing agent policies, blocking malicious prompts, DLP and Conditional Access

    Demo of enforcing agent policies, blocking malicious prompts, applying DLP protections and triggering Conditional Access responses.

Watch on-demand (opens in a new tab)

Resources named in the session

← All sessions