If your AI security plan starts and ends with “LLM,” you’re already behind.

Wil talks with cybersecurity architect Roman Kruglov explains why “AI” is not one risk bucket and how security changes across LLMs, SLMs, RAG, and data-specific models. We talk on-prem vs cloud, API security basics that still get missed, and the bigger problem most teams ignore: data labeling, DLP, and permissions hygiene. You’ll also hear why Copilot can surface what SharePoint search never would, and why identity (including non-human identities) is the next battlefield.

Key takeaways:

• How to compare LLM vs RAG security risks

• Why data labeling + DLP matter before you deploy AI

• Copilot, SharePoint/Teams sprawl, and “public is public”

• Governance templates and guardrails that don’t block delivery

• Why identity is becoming the new perimeter

Follow the show, leave a quick rating, and share this episode with the person pushing “AI” in your org.