Data Security Considerations for AI Adoption
summary
Demonstrates concrete controls: a no‑code sales assistant that queries SharePoint while enforcing Microsoft 365 Copilot RBAC and Purview labels, tenant isolation and encryption, and promises that Copilot data is not used to train foundation models. Shows Azure AI Studio meta‑prompts, input/output content filters and a live jailbreak attempt that the filter blocks. Emphasizes three practical rules — don’t undo existing protections, don’t open apps broadly, and gate prompts — and demos telemetry, transcript and audit workflows.
AI-written summary · about
chapters (11)
- 0:00Introduction to speakers
- 1:03New AI attack surfaces
- 3:15Data challenges in AI
- 6:11Data types and ownership
- 9:51Microsoft 365 security
- 12:51Threat modeling basics
- 17:40Three golden rules of AI
- 23:01Data processing and lifecycle
- 27:47M365 sales assistant demo
- 31:23Azure AI studio demo
- 37:08Final summary and Q&A
related talks
15mAI Guardrails for Enterprise Deployments
10mData Security & AI Adoption: A Leader’s Perspective (Customer Spotlight)
26mCopilot trust & safety: Controls to manage AI risk
01h 11mKeynote - Securing AI: Building Trust in the Era of AI
39mThe Future of Cybersecurity in an AI-driven Era
53mHow AI is impacting cybersecurity. Does it?
21mSecurity in the Age of AI & Copilot
48mBuilding and using AI models responsibly