330 - From vibe-coding to production: Shipping verifiable AI
In this episode, we take a look at the transition from "vibe-coding" to shipping verifiable, production-grade AI applications. This is the critical shift from relying on "gut feel" and prompt tinkering to implementing rigorous audit trails, versioning, and security controls. We discuss why many AI pilots fail due to a lack of explainability, the specific risks of privilege expansion and data leakage in agentic workflows, and also outline how to use Microsoft Foundry/Azure AI Studio to operationalize your models like true regulated software.
(00:00) - Intro and catching up.
(05:30) - Show content starts.
Show links
- RedAmon (GitHub) for automated agentic offensive security
(00:00) - Intro and catching up.
(05:30) - Show content starts.
Show links
- RedAmon (GitHub) for automated agentic offensive security
Creators and Guests
Host
Jussi Roine
Bridging the gap between business and security 🛡 Senior Security Architect at Not Bad Security | Microsoft MVP/MCM
