From Shadow AI to Safe Adoption: Guardrails for Enterprise AI

What's Up with Tech?

Interested in being a guest? Email us at [email protected]

Your chatbot just recommended a competitor. Your interns pasted sensitive research into a public model. And your developers fed entire mobile codebases to a chatbot for “optimization.” We’ve seen it all—and we’re turning those hard lessons into a practical, repeatable playbook for safe AI at scale.
We sit down with Rick Caccia CEO and co‑founder of WitnessAI, to unpack how large organizations move from “Doctor No” to “Doctor Yes.” Rick explains what a confidence layer for enterprise AI looks like: full observability across employee usage, third‑party apps, internal models, customer-facing chatbots, and increasingly powerful agents. We talk about why legacy DLP can’t keep up with conversational risk, how intention-based controls catch unsafe goals in real time, and why brand safety belongs right alongside security and compliance. You’ll hear real stories: flipping 150,000 employees from blocked to safely enabled in days, stopping inadvertent PCI exposure in support workflows, and preventing chatbots from steering customers to competitors.
We also get tactical about regulation and readiness. Yes, the EU AI Act matters—but so do familiar frameworks like PCI DSS and HIPAA that AI usage quietly reactivates. Rick shares a phased roadmap: start with visibility, normalize identity across divisions, roll out targeted policies slowly, and add guardrails that constrain agents before they act. We cover the attacker–defender gap as AI lowers cost and increases speed for adversaries, plus the emerging blind spots leaders should watch as agentic capabilities become default in operating systems and business apps.
If you’re a CISO, CIO, or builder trying to enable AI without losing control, this conversation offers concrete steps, fresh mental models, and a path to say yes with confidence.

Support the show

More at https://linktr.ee/EvanKirstel

פרקים

1. Setting The Stakes: Safe Enterprise AI (00:00:00)

2. Witness AI’s Mission: A Confidence Layer (00:00:53)

3. Defining “Safe” For Employees, Apps, Agents (00:01:18)

4. From Doctor No To Doctor Yes (00:03:12)

5. Shadow AI And The 5,000-App Surprise (00:04:20)

6. Why AI Risk Differs From Traditional Security (00:06:46)

7. Intention-Based Controls Over DLP (00:08:27)

8. Regulations: New Acts And Old Rules (00:09:19)

9. Attackers’ Advantage And New Blind Spots (00:10:40)

10. Real-World Wins: Enabling 150k Safely (00:12:07)

11. B2C Bots And Brand-Safe Guardrails (00:13:46)

12. Playbook: Observe First, Then Control (00:15:16)

13. Closing And Listener CTA (00:18:35)