BRK329 Breakout Thu 19 Nov 16:00 PST
Get your data AI-ready: Discover & mitigate data exposure risks
Shilpa Bothra, Erin Miyake
Summary
From the recording
Speakers argued AI readiness is data readiness. Enhancements to the top 100 out-of-box sets were announced, with the rest to follow over time. Unified classification management is in public preview. Auto-labelling rises from 100K to 500K files, and simulation from 4 million to 20 million. DLP for Copilot grounding using SITs, without labels, was announced at Ignite. Label-based grounding blocks are in market today, as are prompt-text SIT controls. DLP controls for local agents were demonstrated, including Claude Code. Purview alerts from IRM and DLP can go to Defender today. Caveat: captions are auto-generated, and 'out-of-box sets' likely means SITs. Most announced items had no stated availability, licensing or prerequisites.
Session description
Microsoft's catalogue description
As AI accelerates access to information, organizations need stronger protections for sensitive data. Learn how Microsoft Purview combines classification, permissions insights, and controls to reduce oversharing and secure data across AI.
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 16:00
Data exposure risk walkthroughs
Walkthroughs of discovering and mitigating data exposure risks, with further demo points at 17:09.
- Skip to 39:17
DLP policies for agents
The speaker switches to an agent acting and describes DLP policies that can be scoped to all agents or to a specific agent picker.