Skip to content

BRK331 Breakout Thu 19 Nov 13:30 PST

A Single Policy for Humans and AI Agents? Here’s How to Make It Work

Danielle Augustin, Etan Basseri, Sharon Sandhu

Summary

From the recording

Microsoft presented Entra Conditional Access as a single policy engine for people and AI agents. Policy plans, now in public preview, group rules under one objective with shared conditions, with stated limits of 100 plans per tenant and 100 rules per plan. Conditional Access for assistive agents was announced in public preview. Dynamic controls were also announced, allowing, restricting, blocking or requiring human approval at action time. Dynamic agent risk draws on runtime risk, exposure level and blast radius. Captions give the dynamic controls status as 'unlimited preview', which is unclear. Licensing, pricing and GA dates were not stated.

Session description

Microsoft's catalogue description

Dozens of policies become hundreds, and changing one can break something downstream. See how a unified policy engine brings identity, network, device, and risk signals together to make access decisions without forcing teams to untangle dependencies. Learn how to scale adaptive access with confidence and control.​

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 5:51

    Conditional Access policy plans in the Entra portal

    In the Entra portal, creates a "secure developer access" plan, sets shared conditions (include developers, exclude break-glass accounts), and adds a first rule requiring MFA. Limits are 100 plans per tenant and 100 rules per plan.

Watch on-demand (opens in a new tab)

Resources named in the session

← All sessions